source("0__helpers.R")
opts_chunk$set(warning=TRUE, cache=F,cache.lazy=F,tidy=FALSE,autodep=TRUE,dev=c('png','pdf'),fig.width=20,fig.height=12.5,out.width='1440px',out.height='900px',cache.extra=file.info('swed1.rdata')[, 'mtime'])
make_path = function(file) {
get_coefficient_path(file, "krmh")
}
# options for each chunk calling knit_child
opts_chunk$set(warning=FALSE, message = FALSE)
The krmh.1
dataset contains only those participants where paternal age is known, the birthdate is between 1720 and 1850 and the marriage is known (meaning we know when it started and how it ended by spousal death). In known marriages we can assume that missing death dates for the kids mean that they migrated out.
All of the following models are the same as our main model m3, except for the noted changes to test robustness.
For the four historical populations, we imposed quite stringent exclusion criteria to ensure sufficient data quality for our intended analysis. This was not necessary for the modern Swedish data, because there were no exclusion criteria to relax.
model_filename = make_path("r1_relaxed_exclusion_criteria")
if (file.exists(model_filename)) {
cat(summarise_model())
r1 = model
}
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 11848)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2897)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.23 0.01 0.21 0.25 927 1.01
## sd(hu_Intercept) 0.54 0.04 0.47 0.62 874 1.00
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.87 0.28 1.30 2.38 132
## paternalage 0.04 0.05 -0.05 0.14 757
## birth_cohort1670M1700 -0.02 0.31 -0.59 0.62 152
## birth_cohort1700M1720 -0.09 0.29 -0.62 0.52 143
## birth_cohort1720M1760 -0.20 0.28 -0.70 0.37 129
## birth_cohort1760M1765 -0.15 0.28 -0.67 0.42 132
## birth_cohort1765M1770 -0.30 0.28 -0.80 0.28 134
## birth_cohort1770M1775 -0.30 0.28 -0.80 0.27 131
## birth_cohort1775M1780 -0.22 0.28 -0.72 0.36 131
## birth_cohort1780M1785 -0.30 0.28 -0.81 0.26 131
## birth_cohort1785M1790 -0.26 0.28 -0.77 0.32 131
## birth_cohort1790M1795 -0.24 0.28 -0.73 0.34 130
## birth_cohort1795M1800 -0.29 0.28 -0.79 0.28 131
## birth_cohort1800M1805 -0.29 0.28 -0.80 0.28 130
## birth_cohort1805M1810 -0.30 0.28 -0.80 0.27 130
## birth_cohort1810M1815 -0.27 0.28 -0.77 0.30 130
## birth_cohort1815M1820 -0.34 0.28 -0.84 0.24 129
## birth_cohort1820M1825 -0.39 0.28 -0.89 0.18 130
## birth_cohort1825M1830 -0.39 0.28 -0.88 0.19 129
## birth_cohort1830M1835 -0.38 0.28 -0.88 0.19 129
## male1 0.09 0.01 0.06 0.12 3000
## maternalage.factor1420 -0.08 0.08 -0.23 0.07 3000
## maternalage.factor3550 0.01 0.03 -0.04 0.06 3000
## paternalage.mean -0.05 0.05 -0.16 0.04 770
## paternal_loss01 -0.20 0.07 -0.34 -0.06 3000
## paternal_loss15 -0.06 0.05 -0.15 0.04 3000
## paternal_loss510 -0.07 0.04 -0.15 0.01 1793
## paternal_loss1015 0.02 0.04 -0.06 0.09 1724
## paternal_loss1520 -0.08 0.04 -0.15 -0.01 1970
## paternal_loss2025 -0.11 0.03 -0.18 -0.04 2032
## paternal_loss2530 -0.02 0.03 -0.08 0.04 1577
## paternal_loss3035 -0.03 0.03 -0.09 0.03 1593
## paternal_loss3540 -0.01 0.03 -0.07 0.05 1930
## paternal_loss4045 -0.01 0.03 -0.08 0.05 3000
## paternal_lossunclear -0.05 0.04 -0.13 0.04 1813
## maternal_loss01 0.09 0.07 -0.06 0.23 3000
## maternal_loss15 -0.01 0.04 -0.10 0.08 3000
## maternal_loss510 0.06 0.04 -0.02 0.14 1944
## maternal_loss1015 0.03 0.04 -0.05 0.10 3000
## maternal_loss1520 -0.01 0.04 -0.09 0.06 3000
## maternal_loss2025 -0.01 0.04 -0.08 0.07 3000
## maternal_loss2530 -0.04 0.03 -0.11 0.02 3000
## maternal_loss3035 -0.04 0.03 -0.10 0.02 1225
## maternal_loss3540 -0.04 0.03 -0.10 0.01 3000
## maternal_loss4045 -0.01 0.03 -0.07 0.05 3000
## maternal_lossunclear -0.10 0.04 -0.19 -0.02 3000
## older_siblings1 0.03 0.02 -0.02 0.08 1473
## older_siblings2 -0.03 0.03 -0.09 0.04 920
## older_siblings3 -0.06 0.04 -0.14 0.03 791
## older_siblings4 -0.05 0.06 -0.16 0.06 797
## older_siblings5P -0.05 0.07 -0.20 0.08 833
## nr.siblings 0.00 0.01 -0.01 0.02 892
## last_born1 -0.03 0.02 -0.07 0.01 3000
## hu_Intercept -1.22 0.82 -2.82 0.44 129
## hu_paternalage 0.31 0.13 0.05 0.56 983
## hu_birth_cohort1670M1700 1.05 0.90 -0.76 2.89 144
## hu_birth_cohort1700M1720 0.17 0.85 -1.58 1.83 132
## hu_birth_cohort1720M1760 0.83 0.81 -0.80 2.41 123
## hu_birth_cohort1760M1765 0.80 0.82 -0.85 2.41 124
## hu_birth_cohort1765M1770 0.63 0.81 -1.01 2.21 123
## hu_birth_cohort1770M1775 0.77 0.81 -0.87 2.35 124
## hu_birth_cohort1775M1780 0.70 0.81 -0.95 2.29 127
## hu_birth_cohort1780M1785 0.58 0.81 -1.07 2.19 123
## hu_birth_cohort1785M1790 0.45 0.81 -1.18 2.04 123
## hu_birth_cohort1790M1795 0.56 0.81 -1.10 2.17 123
## hu_birth_cohort1795M1800 0.35 0.81 -1.28 1.94 123
## hu_birth_cohort1800M1805 0.26 0.81 -1.42 1.84 124
## hu_birth_cohort1805M1810 0.52 0.81 -1.13 2.10 124
## hu_birth_cohort1810M1815 0.36 0.81 -1.28 1.93 123
## hu_birth_cohort1815M1820 0.11 0.81 -1.55 1.69 124
## hu_birth_cohort1820M1825 0.28 0.81 -1.37 1.87 124
## hu_birth_cohort1825M1830 0.23 0.81 -1.42 1.81 123
## hu_birth_cohort1830M1835 0.23 0.81 -1.43 1.81 123
## hu_male1 0.26 0.04 0.18 0.35 3000
## hu_maternalage.factor1420 0.16 0.20 -0.22 0.56 3000
## hu_maternalage.factor3550 0.10 0.07 -0.03 0.23 3000
## hu_paternalage.mean -0.20 0.13 -0.46 0.06 1049
## hu_paternal_loss01 0.67 0.17 0.33 1.01 3000
## hu_paternal_loss15 0.58 0.12 0.34 0.82 1999
## hu_paternal_loss510 0.25 0.11 0.03 0.46 1983
## hu_paternal_loss1015 0.20 0.10 0.01 0.40 1689
## hu_paternal_loss1520 0.17 0.10 -0.03 0.37 1847
## hu_paternal_loss2025 0.16 0.09 -0.03 0.34 1641
## hu_paternal_loss2530 0.09 0.09 -0.08 0.26 1580
## hu_paternal_loss3035 0.02 0.09 -0.15 0.19 1681
## hu_paternal_loss3540 0.03 0.09 -0.14 0.20 1740
## hu_paternal_loss4045 0.10 0.09 -0.09 0.28 3000
## hu_paternal_lossunclear 0.55 0.11 0.34 0.76 1847
## hu_maternal_loss01 1.60 0.18 1.26 1.94 3000
## hu_maternal_loss15 0.60 0.11 0.38 0.82 3000
## hu_maternal_loss510 0.53 0.10 0.33 0.73 3000
## hu_maternal_loss1015 0.41 0.10 0.20 0.62 3000
## hu_maternal_loss1520 0.33 0.10 0.13 0.54 3000
## hu_maternal_loss2025 0.27 0.10 0.08 0.46 3000
## hu_maternal_loss2530 0.14 0.09 -0.03 0.32 3000
## hu_maternal_loss3035 0.17 0.08 0.02 0.33 3000
## hu_maternal_loss3540 0.09 0.08 -0.06 0.25 3000
## hu_maternal_loss4045 0.28 0.08 0.12 0.43 3000
## hu_maternal_lossunclear 0.86 0.10 0.67 1.06 3000
## hu_older_siblings1 -0.02 0.06 -0.14 0.11 1758
## hu_older_siblings2 -0.16 0.09 -0.33 0.01 1140
## hu_older_siblings3 -0.20 0.11 -0.42 0.03 946
## hu_older_siblings4 -0.23 0.14 -0.51 0.05 954
## hu_older_siblings5P -0.53 0.19 -0.90 -0.17 975
## hu_nr.siblings 0.09 0.02 0.06 0.13 1241
## hu_last_born1 0.05 0.06 -0.06 0.16 3000
## Rhat
## Intercept 1.03
## paternalage 1.00
## birth_cohort1670M1700 1.02
## birth_cohort1700M1720 1.02
## birth_cohort1720M1760 1.03
## birth_cohort1760M1765 1.03
## birth_cohort1765M1770 1.03
## birth_cohort1770M1775 1.03
## birth_cohort1775M1780 1.03
## birth_cohort1780M1785 1.03
## birth_cohort1785M1790 1.03
## birth_cohort1790M1795 1.03
## birth_cohort1795M1800 1.03
## birth_cohort1800M1805 1.03
## birth_cohort1805M1810 1.03
## birth_cohort1810M1815 1.03
## birth_cohort1815M1820 1.03
## birth_cohort1820M1825 1.03
## birth_cohort1825M1830 1.03
## birth_cohort1830M1835 1.03
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.00
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## paternal_lossunclear 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## maternal_lossunclear 1.00
## older_siblings1 1.00
## older_siblings2 1.00
## older_siblings3 1.00
## older_siblings4 1.00
## older_siblings5P 1.00
## nr.siblings 1.00
## last_born1 1.00
## hu_Intercept 1.03
## hu_paternalage 1.00
## hu_birth_cohort1670M1700 1.03
## hu_birth_cohort1700M1720 1.03
## hu_birth_cohort1720M1760 1.03
## hu_birth_cohort1760M1765 1.03
## hu_birth_cohort1765M1770 1.03
## hu_birth_cohort1770M1775 1.03
## hu_birth_cohort1775M1780 1.03
## hu_birth_cohort1780M1785 1.03
## hu_birth_cohort1785M1790 1.03
## hu_birth_cohort1790M1795 1.03
## hu_birth_cohort1795M1800 1.03
## hu_birth_cohort1800M1805 1.03
## hu_birth_cohort1805M1810 1.03
## hu_birth_cohort1810M1815 1.03
## hu_birth_cohort1815M1820 1.03
## hu_birth_cohort1820M1825 1.03
## hu_birth_cohort1825M1830 1.03
## hu_birth_cohort1830M1835 1.03
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.00
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_paternal_lossunclear 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_maternal_lossunclear 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.00
## hu_older_siblings3 1.00
## hu_older_siblings4 1.00
## hu_older_siblings5P 1.00
## hu_nr.siblings 1.00
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 6.462 | 3.656 | 10.8 |
paternalage | 1.041 | 0.9472 | 1.15 |
birth_cohort1670M1700 | 0.9843 | 0.5557 | 1.866 |
birth_cohort1700M1720 | 0.918 | 0.5364 | 1.677 |
birth_cohort1720M1760 | 0.8172 | 0.4943 | 1.452 |
birth_cohort1760M1765 | 0.8579 | 0.5098 | 1.523 |
birth_cohort1765M1770 | 0.7434 | 0.448 | 1.324 |
birth_cohort1770M1775 | 0.7402 | 0.4477 | 1.312 |
birth_cohort1775M1780 | 0.8028 | 0.4858 | 1.43 |
birth_cohort1780M1785 | 0.7383 | 0.446 | 1.301 |
birth_cohort1785M1790 | 0.7701 | 0.463 | 1.373 |
birth_cohort1790M1795 | 0.7895 | 0.4806 | 1.401 |
birth_cohort1795M1800 | 0.7483 | 0.4517 | 1.325 |
birth_cohort1800M1805 | 0.7463 | 0.4513 | 1.326 |
birth_cohort1805M1810 | 0.7419 | 0.4506 | 1.314 |
birth_cohort1810M1815 | 0.7627 | 0.4622 | 1.35 |
birth_cohort1815M1820 | 0.709 | 0.4306 | 1.276 |
birth_cohort1820M1825 | 0.6803 | 0.4118 | 1.198 |
birth_cohort1825M1830 | 0.6802 | 0.4138 | 1.207 |
birth_cohort1830M1835 | 0.6832 | 0.4161 | 1.207 |
male1 | 1.09 | 1.058 | 1.122 |
maternalage.factor1420 | 0.9207 | 0.7912 | 1.073 |
maternalage.factor3550 | 1.009 | 0.9601 | 1.061 |
paternalage.mean | 0.9485 | 0.8559 | 1.045 |
paternal_loss01 | 0.8212 | 0.7138 | 0.9382 |
paternal_loss15 | 0.946 | 0.8604 | 1.038 |
paternal_loss510 | 0.9311 | 0.8609 | 1.009 |
paternal_loss1015 | 1.016 | 0.9422 | 1.095 |
paternal_loss1520 | 0.9229 | 0.8588 | 0.9902 |
paternal_loss2025 | 0.8948 | 0.8365 | 0.9583 |
paternal_loss2530 | 0.9797 | 0.9206 | 1.04 |
paternal_loss3035 | 0.9689 | 0.9137 | 1.028 |
paternal_loss3540 | 0.9901 | 0.9326 | 1.052 |
paternal_loss4045 | 0.9861 | 0.9224 | 1.054 |
paternal_lossunclear | 0.9542 | 0.8802 | 1.036 |
maternal_loss01 | 1.098 | 0.9458 | 1.265 |
maternal_loss15 | 0.9873 | 0.9024 | 1.079 |
maternal_loss510 | 1.058 | 0.977 | 1.145 |
maternal_loss1015 | 1.029 | 0.9531 | 1.11 |
maternal_loss1520 | 0.9905 | 0.9156 | 1.066 |
maternal_loss2025 | 0.9941 | 0.9257 | 1.067 |
maternal_loss2530 | 0.9581 | 0.8984 | 1.024 |
maternal_loss3035 | 0.9603 | 0.9041 | 1.02 |
maternal_loss3540 | 0.9562 | 0.9056 | 1.008 |
maternal_loss4045 | 0.9898 | 0.9327 | 1.05 |
maternal_lossunclear | 0.9023 | 0.8299 | 0.983 |
older_siblings1 | 1.033 | 0.9837 | 1.084 |
older_siblings2 | 0.9739 | 0.9126 | 1.04 |
older_siblings3 | 0.9452 | 0.8728 | 1.027 |
older_siblings4 | 0.9524 | 0.855 | 1.06 |
older_siblings5P | 0.9474 | 0.8209 | 1.088 |
nr.siblings | 1.005 | 0.9923 | 1.018 |
last_born1 | 0.9699 | 0.9293 | 1.01 |
hu_Intercept | 0.2954 | 0.05966 | 1.55 |
hu_paternalage | 1.364 | 1.054 | 1.754 |
hu_birth_cohort1670M1700 | 2.869 | 0.4688 | 18.03 |
hu_birth_cohort1700M1720 | 1.188 | 0.2061 | 6.237 |
hu_birth_cohort1720M1760 | 2.299 | 0.4491 | 11.16 |
hu_birth_cohort1760M1765 | 2.235 | 0.4276 | 11.11 |
hu_birth_cohort1765M1770 | 1.877 | 0.3653 | 9.075 |
hu_birth_cohort1770M1775 | 2.17 | 0.4184 | 10.53 |
hu_birth_cohort1775M1780 | 2.006 | 0.3881 | 9.866 |
hu_birth_cohort1780M1785 | 1.793 | 0.3421 | 8.924 |
hu_birth_cohort1785M1790 | 1.567 | 0.3061 | 7.7 |
hu_birth_cohort1790M1795 | 1.749 | 0.3319 | 8.801 |
hu_birth_cohort1795M1800 | 1.421 | 0.2779 | 6.976 |
hu_birth_cohort1800M1805 | 1.293 | 0.2413 | 6.319 |
hu_birth_cohort1805M1810 | 1.685 | 0.322 | 8.167 |
hu_birth_cohort1810M1815 | 1.427 | 0.277 | 6.884 |
hu_birth_cohort1815M1820 | 1.117 | 0.212 | 5.436 |
hu_birth_cohort1820M1825 | 1.326 | 0.2551 | 6.488 |
hu_birth_cohort1825M1830 | 1.255 | 0.242 | 6.112 |
hu_birth_cohort1830M1835 | 1.253 | 0.2395 | 6.086 |
hu_male1 | 1.303 | 1.199 | 1.418 |
hu_maternalage.factor1420 | 1.175 | 0.8043 | 1.753 |
hu_maternalage.factor3550 | 1.104 | 0.9676 | 1.261 |
hu_paternalage.mean | 0.8184 | 0.6344 | 1.063 |
hu_paternal_loss01 | 1.955 | 1.397 | 2.75 |
hu_paternal_loss15 | 1.783 | 1.399 | 2.265 |
hu_paternal_loss510 | 1.284 | 1.034 | 1.592 |
hu_paternal_loss1015 | 1.225 | 1.014 | 1.498 |
hu_paternal_loss1520 | 1.186 | 0.9735 | 1.444 |
hu_paternal_loss2025 | 1.169 | 0.9723 | 1.41 |
hu_paternal_loss2530 | 1.097 | 0.9223 | 1.297 |
hu_paternal_loss3035 | 1.022 | 0.8589 | 1.214 |
hu_paternal_loss3540 | 1.028 | 0.8698 | 1.22 |
hu_paternal_loss4045 | 1.107 | 0.917 | 1.328 |
hu_paternal_lossunclear | 1.741 | 1.408 | 2.138 |
hu_maternal_loss01 | 4.954 | 3.534 | 6.975 |
hu_maternal_loss15 | 1.82 | 1.457 | 2.262 |
hu_maternal_loss510 | 1.698 | 1.39 | 2.072 |
hu_maternal_loss1015 | 1.51 | 1.224 | 1.856 |
hu_maternal_loss1520 | 1.395 | 1.144 | 1.714 |
hu_maternal_loss2025 | 1.31 | 1.084 | 1.577 |
hu_maternal_loss2530 | 1.147 | 0.9662 | 1.371 |
hu_maternal_loss3035 | 1.188 | 1.016 | 1.394 |
hu_maternal_loss3540 | 1.096 | 0.9445 | 1.278 |
hu_maternal_loss4045 | 1.318 | 1.126 | 1.542 |
hu_maternal_lossunclear | 2.368 | 1.945 | 2.88 |
hu_older_siblings1 | 0.9846 | 0.8684 | 1.116 |
hu_older_siblings2 | 0.8558 | 0.7187 | 1.012 |
hu_older_siblings3 | 0.8201 | 0.6551 | 1.034 |
hu_older_siblings4 | 0.7952 | 0.601 | 1.048 |
hu_older_siblings5P | 0.5874 | 0.4078 | 0.8462 |
hu_nr.siblings | 1.099 | 1.063 | 1.136 |
hu_last_born1 | 1.049 | 0.9372 | 1.177 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 3.95 | [1.55;7.6] | [2.23;6.08] |
estimate father 35y | 3.65 | [1.3;7.57] | [1.94;5.82] |
percentage change | -7.11 | [-23.86;6.73] | [-18.09;2.42] |
OR/IRR | 1.04 | [0.95;1.15] | [0.98;1.11] |
OR hurdle | 1.36 | [1.05;1.75] | [1.15;1.61] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r1_relaxed_exclusion_criteria.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
Adding covariates increases the complexity of the model and makes it harder to interpret. We chose to adjust for many potential confounds because we are interested in causal isolation of the paternal age effect. Here we show what happens when only birth cohort and average paternal age in the family are adjusted for.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + paternalage.mean + (1 | idParents)
## hu ~ paternalage + birth_cohort + paternalage.mean + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.23 0.01 0.20 0.25 1258 1
## sd(hu_Intercept) 0.50 0.04 0.42 0.58 1017 1
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.70 0.07 1.56 1.83 1587
## paternalage -0.03 0.02 -0.06 0.00 3000
## birth_cohort1760M1765 0.00 0.06 -0.12 0.13 3000
## birth_cohort1765M1770 -0.11 0.06 -0.22 -0.01 1284
## birth_cohort1770M1775 -0.11 0.06 -0.23 0.00 1150
## birth_cohort1775M1780 -0.03 0.06 -0.13 0.08 987
## birth_cohort1780M1785 -0.11 0.06 -0.22 0.00 1058
## birth_cohort1785M1790 -0.10 0.06 -0.21 0.01 996
## birth_cohort1790M1795 -0.08 0.05 -0.18 0.02 956
## birth_cohort1795M1800 -0.10 0.05 -0.20 0.00 910
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.01 831
## birth_cohort1805M1810 -0.13 0.05 -0.23 -0.03 927
## birth_cohort1810M1815 -0.10 0.05 -0.19 0.00 843
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 800
## birth_cohort1820M1825 -0.19 0.05 -0.27 -0.09 799
## birth_cohort1825M1830 -0.22 0.05 -0.31 -0.13 793
## birth_cohort1830M1835 -0.19 0.05 -0.29 -0.10 878
## paternalage.mean 0.01 0.02 -0.03 0.06 3000
## hu_Intercept 0.64 0.17 0.30 0.98 1615
## hu_paternalage 0.22 0.05 0.13 0.31 3000
## hu_birth_cohort1760M1765 -0.06 0.17 -0.38 0.28 3000
## hu_birth_cohort1765M1770 -0.35 0.15 -0.63 -0.06 1419
## hu_birth_cohort1770M1775 -0.16 0.15 -0.44 0.13 1263
## hu_birth_cohort1775M1780 -0.25 0.15 -0.54 0.04 1305
## hu_birth_cohort1780M1785 -0.32 0.15 -0.60 -0.03 1274
## hu_birth_cohort1785M1790 -0.49 0.15 -0.78 -0.21 1297
## hu_birth_cohort1790M1795 -0.40 0.14 -0.67 -0.14 1129
## hu_birth_cohort1795M1800 -0.55 0.13 -0.81 -0.31 1057
## hu_birth_cohort1800M1805 -0.64 0.13 -0.90 -0.39 1021
## hu_birth_cohort1805M1810 -0.37 0.13 -0.61 -0.12 999
## hu_birth_cohort1810M1815 -0.53 0.12 -0.77 -0.29 1013
## hu_birth_cohort1815M1820 -0.81 0.12 -1.04 -0.57 1004
## hu_birth_cohort1820M1825 -0.67 0.12 -0.90 -0.44 971
## hu_birth_cohort1825M1830 -0.66 0.12 -0.88 -0.43 953
## hu_birth_cohort1830M1835 -0.66 0.12 -0.90 -0.43 936
## hu_paternalage.mean -0.14 0.06 -0.25 -0.02 3000
## Rhat
## Intercept 1.00
## paternalage 1.00
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.00
## birth_cohort1770M1775 1.00
## birth_cohort1775M1780 1.01
## birth_cohort1780M1785 1.00
## birth_cohort1785M1790 1.00
## birth_cohort1790M1795 1.00
## birth_cohort1795M1800 1.00
## birth_cohort1800M1805 1.00
## birth_cohort1805M1810 1.00
## birth_cohort1810M1815 1.00
## birth_cohort1815M1820 1.00
## birth_cohort1820M1825 1.00
## birth_cohort1825M1830 1.00
## birth_cohort1830M1835 1.00
## paternalage.mean 1.00
## hu_Intercept 1.00
## hu_paternalage 1.00
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.00
## hu_birth_cohort1775M1780 1.00
## hu_birth_cohort1780M1785 1.01
## hu_birth_cohort1785M1790 1.00
## hu_birth_cohort1790M1795 1.01
## hu_birth_cohort1795M1800 1.00
## hu_birth_cohort1800M1805 1.01
## hu_birth_cohort1805M1810 1.01
## hu_birth_cohort1810M1815 1.01
## hu_birth_cohort1815M1820 1.01
## hu_birth_cohort1820M1825 1.01
## hu_birth_cohort1825M1830 1.01
## hu_birth_cohort1830M1835 1.01
## hu_paternalage.mean 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.449 | 4.754 | 6.234 |
paternalage | 0.9717 | 0.9395 | 1.005 |
birth_cohort1760M1765 | 1.004 | 0.8848 | 1.141 |
birth_cohort1765M1770 | 0.8918 | 0.799 | 0.9949 |
birth_cohort1770M1775 | 0.8937 | 0.7965 | 1.001 |
birth_cohort1775M1780 | 0.9733 | 0.875 | 1.084 |
birth_cohort1780M1785 | 0.8931 | 0.7992 | 0.9993 |
birth_cohort1785M1790 | 0.9041 | 0.8105 | 1.011 |
birth_cohort1790M1795 | 0.9219 | 0.8335 | 1.021 |
birth_cohort1795M1800 | 0.9021 | 0.8188 | 0.9956 |
birth_cohort1800M1805 | 0.894 | 0.8133 | 0.9888 |
birth_cohort1805M1810 | 0.8772 | 0.7972 | 0.9666 |
birth_cohort1810M1815 | 0.9083 | 0.8285 | 0.996 |
birth_cohort1815M1820 | 0.8685 | 0.7959 | 0.951 |
birth_cohort1820M1825 | 0.8303 | 0.7606 | 0.9099 |
birth_cohort1825M1830 | 0.8026 | 0.7331 | 0.8794 |
birth_cohort1830M1835 | 0.8231 | 0.7465 | 0.9014 |
paternalage.mean | 1.013 | 0.9663 | 1.062 |
hu_Intercept | 1.904 | 1.346 | 2.671 |
hu_paternalage | 1.244 | 1.136 | 1.357 |
hu_birth_cohort1760M1765 | 0.9459 | 0.681 | 1.329 |
hu_birth_cohort1765M1770 | 0.7061 | 0.53 | 0.9461 |
hu_birth_cohort1770M1775 | 0.8558 | 0.6418 | 1.144 |
hu_birth_cohort1775M1780 | 0.7783 | 0.585 | 1.041 |
hu_birth_cohort1780M1785 | 0.7278 | 0.5478 | 0.9658 |
hu_birth_cohort1785M1790 | 0.6103 | 0.458 | 0.8125 |
hu_birth_cohort1790M1795 | 0.6717 | 0.5103 | 0.8676 |
hu_birth_cohort1795M1800 | 0.5744 | 0.4468 | 0.7358 |
hu_birth_cohort1800M1805 | 0.5292 | 0.4084 | 0.6766 |
hu_birth_cohort1805M1810 | 0.6937 | 0.5435 | 0.8855 |
hu_birth_cohort1810M1815 | 0.5878 | 0.4644 | 0.7451 |
hu_birth_cohort1815M1820 | 0.4461 | 0.353 | 0.5634 |
hu_birth_cohort1820M1825 | 0.5134 | 0.407 | 0.6458 |
hu_birth_cohort1825M1830 | 0.5176 | 0.4157 | 0.6499 |
hu_birth_cohort1830M1835 | 0.5148 | 0.4057 | 0.6522 |
hu_paternalage.mean | 0.8712 | 0.776 | 0.9808 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 1.75 | [1.51;2.01] | [1.59;1.92] |
estimate father 35y | 1.47 | [1.27;1.69] | [1.33;1.61] |
percentage change | -16.37 | [-22.17;-10.38] | [-20.15;-12.42] |
OR/IRR | 0.97 | [0.94;1] | [0.95;0.99] |
OR hurdle | 1.24 | [1.14;1.36] | [1.17;1.32] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r2_few_controls.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
We chose to control for birth order/number of older siblings as a categorical variable, lumping all those who had more than 5 in the category 5+. Because a continuous covariate is also plausible, we tested this alternative model as well.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 1000; warmup = 500; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.22 0.01 0.20 0.25 983 1.00
## sd(hu_Intercept) 0.47 0.05 0.38 0.56 606 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.67 0.08 1.51 1.82 727
## paternalage 0.03 0.06 -0.09 0.16 1001
## birth_cohort1760M1765 0.00 0.06 -0.13 0.13 904
## birth_cohort1765M1770 -0.12 0.06 -0.23 -0.01 671
## birth_cohort1770M1775 -0.11 0.06 -0.23 0.00 619
## birth_cohort1775M1780 -0.02 0.06 -0.13 0.09 590
## birth_cohort1780M1785 -0.11 0.06 -0.23 0.00 638
## birth_cohort1785M1790 -0.09 0.05 -0.20 0.01 585
## birth_cohort1790M1795 -0.08 0.05 -0.18 0.02 485
## birth_cohort1795M1800 -0.11 0.05 -0.20 -0.01 502
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.01 487
## birth_cohort1805M1810 -0.13 0.05 -0.24 -0.03 480
## birth_cohort1810M1815 -0.10 0.05 -0.19 0.00 433
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 441
## birth_cohort1820M1825 -0.19 0.05 -0.28 -0.09 400
## birth_cohort1825M1830 -0.21 0.05 -0.31 -0.11 434
## birth_cohort1830M1835 -0.18 0.05 -0.28 -0.09 411
## male1 0.08 0.02 0.05 0.11 3000
## maternalage.factor1420 -0.06 0.09 -0.25 0.11 3000
## maternalage.factor3550 0.00 0.03 -0.05 0.06 3000
## paternalage.mean -0.03 0.06 -0.16 0.09 993
## paternal_loss01 -0.15 0.07 -0.29 0.00 3000
## paternal_loss15 -0.03 0.05 -0.13 0.06 1685
## paternal_loss510 -0.07 0.04 -0.15 0.02 1626
## paternal_loss1015 0.01 0.04 -0.07 0.08 1263
## paternal_loss1520 -0.09 0.04 -0.17 -0.02 1355
## paternal_loss2025 -0.12 0.04 -0.19 -0.04 1112
## paternal_loss2530 -0.01 0.03 -0.08 0.06 1145
## paternal_loss3035 -0.03 0.03 -0.09 0.04 1219
## paternal_loss3540 -0.01 0.03 -0.07 0.05 1054
## paternal_loss4045 -0.01 0.04 -0.08 0.06 1429
## maternal_loss01 0.10 0.08 -0.05 0.25 3000
## maternal_loss15 -0.02 0.05 -0.11 0.07 1717
## maternal_loss510 0.07 0.04 -0.01 0.14 1424
## maternal_loss1015 0.03 0.04 -0.06 0.11 1526
## maternal_loss1520 0.01 0.04 -0.07 0.08 3000
## maternal_loss2025 0.01 0.04 -0.07 0.08 1343
## maternal_loss2530 -0.02 0.03 -0.09 0.04 1280
## maternal_loss3035 -0.05 0.03 -0.12 0.01 1241
## maternal_loss3540 -0.03 0.03 -0.09 0.02 1273
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## older_siblings -0.01 0.02 -0.04 0.02 1264
## nr.siblings 0.01 0.01 -0.01 0.02 1357
## last_born1 -0.05 0.02 -0.09 0.00 3000
## hu_Intercept -0.14 0.20 -0.54 0.28 861
## hu_paternalage 0.57 0.18 0.22 0.92 963
## hu_birth_cohort1760M1765 -0.03 0.17 -0.35 0.31 1044
## hu_birth_cohort1765M1770 -0.30 0.15 -0.58 -0.02 678
## hu_birth_cohort1770M1775 -0.05 0.15 -0.35 0.22 645
## hu_birth_cohort1775M1780 -0.18 0.14 -0.45 0.10 653
## hu_birth_cohort1780M1785 -0.26 0.15 -0.56 0.01 672
## hu_birth_cohort1785M1790 -0.42 0.14 -0.70 -0.13 722
## hu_birth_cohort1790M1795 -0.30 0.14 -0.56 -0.05 607
## hu_birth_cohort1795M1800 -0.45 0.13 -0.70 -0.20 569
## hu_birth_cohort1800M1805 -0.52 0.13 -0.78 -0.27 627
## hu_birth_cohort1805M1810 -0.25 0.13 -0.50 0.00 530
## hu_birth_cohort1810M1815 -0.42 0.12 -0.67 -0.17 493
## hu_birth_cohort1815M1820 -0.69 0.12 -0.92 -0.46 507
## hu_birth_cohort1820M1825 -0.51 0.12 -0.73 -0.28 497
## hu_birth_cohort1825M1830 -0.53 0.12 -0.76 -0.30 537
## hu_birth_cohort1830M1835 -0.55 0.12 -0.78 -0.31 525
## hu_male1 0.27 0.05 0.18 0.37 3000
## hu_maternalage.factor1420 0.20 0.23 -0.28 0.65 3000
## hu_maternalage.factor3550 0.11 0.07 -0.03 0.25 3000
## hu_paternalage.mean -0.49 0.18 -0.84 -0.13 1000
## hu_paternal_loss01 0.58 0.19 0.21 0.94 3000
## hu_paternal_loss15 0.52 0.14 0.26 0.80 1286
## hu_paternal_loss510 0.19 0.11 -0.04 0.41 1257
## hu_paternal_loss1015 0.15 0.11 -0.06 0.36 1022
## hu_paternal_loss1520 0.10 0.10 -0.10 0.30 957
## hu_paternal_loss2025 0.16 0.10 -0.04 0.36 1022
## hu_paternal_loss2530 0.06 0.10 -0.13 0.24 900
## hu_paternal_loss3035 -0.02 0.09 -0.20 0.17 1009
## hu_paternal_loss3540 -0.01 0.09 -0.19 0.17 984
## hu_paternal_loss4045 0.14 0.10 -0.05 0.34 1244
## hu_maternal_loss01 1.58 0.19 1.23 1.96 3000
## hu_maternal_loss15 0.58 0.12 0.34 0.82 3000
## hu_maternal_loss510 0.48 0.11 0.27 0.68 3000
## hu_maternal_loss1015 0.47 0.11 0.27 0.69 3000
## hu_maternal_loss1520 0.31 0.11 0.10 0.52 3000
## hu_maternal_loss2025 0.25 0.10 0.06 0.45 3000
## hu_maternal_loss2530 0.19 0.09 0.00 0.37 3000
## hu_maternal_loss3035 0.21 0.08 0.05 0.38 1796
## hu_maternal_loss3540 0.07 0.08 -0.09 0.23 1644
## hu_maternal_loss4045 0.27 0.08 0.11 0.44 3000
## hu_older_siblings -0.17 0.04 -0.25 -0.08 990
## hu_nr.siblings 0.14 0.02 0.10 0.19 1040
## hu_last_born1 0.09 0.06 -0.04 0.22 3000
## Rhat
## Intercept 1.00
## paternalage 1.01
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.00
## birth_cohort1770M1775 1.00
## birth_cohort1775M1780 1.00
## birth_cohort1780M1785 1.00
## birth_cohort1785M1790 1.00
## birth_cohort1790M1795 1.00
## birth_cohort1795M1800 1.00
## birth_cohort1800M1805 1.01
## birth_cohort1805M1810 1.00
## birth_cohort1810M1815 1.00
## birth_cohort1815M1820 1.00
## birth_cohort1820M1825 1.00
## birth_cohort1825M1830 1.00
## birth_cohort1830M1835 1.01
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.01
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings 1.01
## nr.siblings 1.00
## last_born1 1.00
## hu_Intercept 1.00
## hu_paternalage 1.00
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.01
## hu_birth_cohort1770M1775 1.01
## hu_birth_cohort1775M1780 1.00
## hu_birth_cohort1780M1785 1.01
## hu_birth_cohort1785M1790 1.00
## hu_birth_cohort1790M1795 1.01
## hu_birth_cohort1795M1800 1.01
## hu_birth_cohort1800M1805 1.01
## hu_birth_cohort1805M1810 1.01
## hu_birth_cohort1810M1815 1.01
## hu_birth_cohort1815M1820 1.01
## hu_birth_cohort1820M1825 1.01
## hu_birth_cohort1825M1830 1.01
## hu_birth_cohort1830M1835 1.01
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.00
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.01
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_older_siblings 1.00
## hu_nr.siblings 1.00
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.3 | 4.522 | 6.182 |
paternalage | 1.032 | 0.9124 | 1.168 |
birth_cohort1760M1765 | 0.9996 | 0.8797 | 1.135 |
birth_cohort1765M1770 | 0.8879 | 0.7921 | 0.9918 |
birth_cohort1770M1775 | 0.8929 | 0.7941 | 0.9984 |
birth_cohort1775M1780 | 0.9767 | 0.8739 | 1.089 |
birth_cohort1780M1785 | 0.8931 | 0.7929 | 1.003 |
birth_cohort1785M1790 | 0.9108 | 0.818 | 1.01 |
birth_cohort1790M1795 | 0.9249 | 0.8352 | 1.025 |
birth_cohort1795M1800 | 0.9003 | 0.8176 | 0.9931 |
birth_cohort1800M1805 | 0.8955 | 0.8136 | 0.9874 |
birth_cohort1805M1810 | 0.8742 | 0.7899 | 0.97 |
birth_cohort1810M1815 | 0.9062 | 0.8248 | 1.001 |
birth_cohort1815M1820 | 0.8691 | 0.7939 | 0.9556 |
birth_cohort1820M1825 | 0.8308 | 0.7567 | 0.9148 |
birth_cohort1825M1830 | 0.8113 | 0.7366 | 0.8925 |
birth_cohort1830M1835 | 0.8337 | 0.7571 | 0.9169 |
male1 | 1.082 | 1.047 | 1.118 |
maternalage.factor1420 | 0.9379 | 0.7825 | 1.118 |
maternalage.factor3550 | 1.004 | 0.9511 | 1.057 |
paternalage.mean | 0.9671 | 0.854 | 1.097 |
paternal_loss01 | 0.865 | 0.7492 | 0.9973 |
paternal_loss15 | 0.9661 | 0.8747 | 1.065 |
paternal_loss510 | 0.9362 | 0.8627 | 1.019 |
paternal_loss1015 | 1.006 | 0.9323 | 1.088 |
paternal_loss1520 | 0.9106 | 0.8458 | 0.9825 |
paternal_loss2025 | 0.8883 | 0.8249 | 0.9572 |
paternal_loss2530 | 0.9904 | 0.9254 | 1.059 |
paternal_loss3035 | 0.9724 | 0.9111 | 1.036 |
paternal_loss3540 | 0.9896 | 0.9311 | 1.055 |
paternal_loss4045 | 0.9917 | 0.926 | 1.064 |
maternal_loss01 | 1.105 | 0.949 | 1.284 |
maternal_loss15 | 0.9826 | 0.896 | 1.074 |
maternal_loss510 | 1.069 | 0.9871 | 1.154 |
maternal_loss1015 | 1.027 | 0.9461 | 1.111 |
maternal_loss1520 | 1.005 | 0.9315 | 1.085 |
maternal_loss2025 | 1.008 | 0.933 | 1.084 |
maternal_loss2530 | 0.9779 | 0.9129 | 1.046 |
maternal_loss3035 | 0.9467 | 0.886 | 1.01 |
maternal_loss3540 | 0.9663 | 0.912 | 1.023 |
maternal_loss4045 | 0.971 | 0.9095 | 1.032 |
older_siblings | 0.9893 | 0.9599 | 1.019 |
nr.siblings | 1.006 | 0.9894 | 1.024 |
last_born1 | 0.9559 | 0.9117 | 1.004 |
hu_Intercept | 0.8698 | 0.585 | 1.32 |
hu_paternalage | 1.77 | 1.25 | 2.498 |
hu_birth_cohort1760M1765 | 0.9697 | 0.7023 | 1.359 |
hu_birth_cohort1765M1770 | 0.7406 | 0.5604 | 0.9838 |
hu_birth_cohort1770M1775 | 0.9487 | 0.7073 | 1.252 |
hu_birth_cohort1775M1780 | 0.8317 | 0.6359 | 1.107 |
hu_birth_cohort1780M1785 | 0.7673 | 0.5722 | 1.012 |
hu_birth_cohort1785M1790 | 0.6593 | 0.4979 | 0.8792 |
hu_birth_cohort1790M1795 | 0.7383 | 0.573 | 0.9544 |
hu_birth_cohort1795M1800 | 0.6389 | 0.4967 | 0.8198 |
hu_birth_cohort1800M1805 | 0.593 | 0.4603 | 0.7629 |
hu_birth_cohort1805M1810 | 0.7798 | 0.6087 | 1 |
hu_birth_cohort1810M1815 | 0.6544 | 0.5137 | 0.8404 |
hu_birth_cohort1815M1820 | 0.5019 | 0.3977 | 0.6316 |
hu_birth_cohort1820M1825 | 0.6015 | 0.4824 | 0.7589 |
hu_birth_cohort1825M1830 | 0.5888 | 0.4658 | 0.7435 |
hu_birth_cohort1830M1835 | 0.5772 | 0.4597 | 0.7303 |
hu_male1 | 1.316 | 1.201 | 1.444 |
hu_maternalage.factor1420 | 1.217 | 0.759 | 1.92 |
hu_maternalage.factor3550 | 1.113 | 0.9658 | 1.282 |
hu_paternalage.mean | 0.6129 | 0.4323 | 0.8774 |
hu_paternal_loss01 | 1.784 | 1.238 | 2.557 |
hu_paternal_loss15 | 1.688 | 1.302 | 2.218 |
hu_paternal_loss510 | 1.208 | 0.9583 | 1.502 |
hu_paternal_loss1015 | 1.165 | 0.9437 | 1.432 |
hu_paternal_loss1520 | 1.105 | 0.9048 | 1.348 |
hu_paternal_loss2025 | 1.173 | 0.9621 | 1.437 |
hu_paternal_loss2530 | 1.061 | 0.8751 | 1.277 |
hu_paternal_loss3035 | 0.982 | 0.8155 | 1.187 |
hu_paternal_loss3540 | 0.991 | 0.8237 | 1.189 |
hu_paternal_loss4045 | 1.154 | 0.9509 | 1.408 |
hu_maternal_loss01 | 4.852 | 3.405 | 7.07 |
hu_maternal_loss15 | 1.781 | 1.411 | 2.261 |
hu_maternal_loss510 | 1.612 | 1.309 | 1.977 |
hu_maternal_loss1015 | 1.6 | 1.309 | 1.994 |
hu_maternal_loss1520 | 1.367 | 1.107 | 1.687 |
hu_maternal_loss2025 | 1.289 | 1.057 | 1.566 |
hu_maternal_loss2530 | 1.205 | 1.005 | 1.442 |
hu_maternal_loss3035 | 1.237 | 1.052 | 1.463 |
hu_maternal_loss3540 | 1.068 | 0.9125 | 1.254 |
hu_maternal_loss4045 | 1.312 | 1.115 | 1.549 |
hu_older_siblings | 0.8472 | 0.7797 | 0.9202 |
hu_nr.siblings | 1.155 | 1.105 | 1.21 |
hu_last_born1 | 1.095 | 0.9649 | 1.242 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.78 | [2.26;3.38] | [2.43;3.18] |
estimate father 35y | 2.12 | [1.78;2.52] | [1.89;2.39] |
percentage change | -23.54 | [-37.92;-5.4] | [-33.32;-12.25] |
OR/IRR | 1.03 | [0.91;1.17] | [0.95;1.12] |
OR hurdle | 1.77 | [1.25;2.5] | [1.41;2.23] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r3_birth_order_continuous.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
Birth order is usually used as a proxy variable for parental investment, the assumption being that older siblings require parental attention. However, there are are reasons to doubt this, as fully-grown siblings probably do not compete for the same resources. To compute a clearer proxy variable of competing siblings, we computed and adjusted for the number of siblings who were alive and younger than five at the time of birth of the anchor child.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + nr.siblings + dependent_sibs_f5y + (1 | idParents)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + nr.siblings + dependent_sibs_f5y + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 1000; warmup = 500; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.23 0.01 0.20 0.25 1126 1.00
## sd(hu_Intercept) 0.52 0.04 0.43 0.60 571 1.02
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.65 0.08 1.50 1.80 1208
## paternalage -0.02 0.03 -0.07 0.03 1156
## birth_cohort1760M1765 0.00 0.06 -0.13 0.12 1145
## birth_cohort1765M1770 -0.12 0.06 -0.23 -0.01 872
## birth_cohort1770M1775 -0.11 0.06 -0.22 0.00 945
## birth_cohort1775M1780 -0.03 0.06 -0.14 0.09 868
## birth_cohort1780M1785 -0.11 0.06 -0.22 0.00 828
## birth_cohort1785M1790 -0.09 0.06 -0.20 0.01 844
## birth_cohort1790M1795 -0.08 0.05 -0.18 0.03 778
## birth_cohort1795M1800 -0.10 0.05 -0.20 -0.01 766
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.02 704
## birth_cohort1805M1810 -0.13 0.05 -0.23 -0.04 747
## birth_cohort1810M1815 -0.10 0.05 -0.19 0.00 664
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 632
## birth_cohort1820M1825 -0.18 0.05 -0.28 -0.09 653
## birth_cohort1825M1830 -0.21 0.05 -0.30 -0.12 638
## birth_cohort1830M1835 -0.18 0.05 -0.28 -0.09 690
## male1 0.08 0.02 0.05 0.11 3000
## maternalage.factor1420 -0.07 0.09 -0.26 0.11 3000
## maternalage.factor3550 0.00 0.03 -0.06 0.05 1711
## paternalage.mean 0.01 0.03 -0.04 0.08 1505
## paternal_loss01 -0.14 0.07 -0.28 0.01 3000
## paternal_loss15 -0.04 0.05 -0.14 0.06 1692
## paternal_loss510 -0.07 0.04 -0.15 0.01 1360
## paternal_loss1015 0.00 0.04 -0.07 0.08 1554
## paternal_loss1520 -0.10 0.04 -0.17 -0.02 1068
## paternal_loss2025 -0.12 0.04 -0.19 -0.05 1244
## paternal_loss2530 -0.01 0.03 -0.08 0.05 819
## paternal_loss3035 -0.03 0.03 -0.10 0.03 1299
## paternal_loss3540 -0.01 0.03 -0.08 0.05 1417
## paternal_loss4045 -0.01 0.04 -0.08 0.06 1546
## maternal_loss01 0.10 0.08 -0.05 0.26 3000
## maternal_loss15 -0.03 0.05 -0.12 0.06 3000
## maternal_loss510 0.07 0.04 -0.01 0.14 1966
## maternal_loss1015 0.03 0.04 -0.05 0.11 1842
## maternal_loss1520 0.01 0.04 -0.07 0.09 2108
## maternal_loss2025 0.01 0.04 -0.07 0.08 1939
## maternal_loss2530 -0.02 0.03 -0.09 0.04 1768
## maternal_loss3035 -0.05 0.03 -0.12 0.01 2069
## maternal_loss3540 -0.03 0.03 -0.09 0.02 1787
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## nr.siblings 0.00 0.00 -0.01 0.01 3000
## dependent_sibs_f5y 0.00 0.01 -0.02 0.02 3000
## hu_Intercept -0.41 0.20 -0.80 -0.01 1009
## hu_paternalage -0.04 0.07 -0.18 0.10 1546
## hu_birth_cohort1760M1765 -0.03 0.17 -0.35 0.30 1203
## hu_birth_cohort1765M1770 -0.31 0.15 -0.60 -0.02 887
## hu_birth_cohort1770M1775 -0.07 0.15 -0.35 0.22 722
## hu_birth_cohort1775M1780 -0.19 0.15 -0.47 0.10 668
## hu_birth_cohort1780M1785 -0.25 0.16 -0.55 0.05 774
## hu_birth_cohort1785M1790 -0.41 0.15 -0.70 -0.13 750
## hu_birth_cohort1790M1795 -0.30 0.14 -0.57 -0.03 655
## hu_birth_cohort1795M1800 -0.47 0.13 -0.72 -0.22 597
## hu_birth_cohort1800M1805 -0.53 0.13 -0.79 -0.29 589
## hu_birth_cohort1805M1810 -0.26 0.13 -0.51 -0.01 596
## hu_birth_cohort1810M1815 -0.43 0.13 -0.68 -0.20 606
## hu_birth_cohort1815M1820 -0.71 0.12 -0.95 -0.47 565
## hu_birth_cohort1820M1825 -0.54 0.12 -0.78 -0.30 568
## hu_birth_cohort1825M1830 -0.54 0.12 -0.78 -0.31 575
## hu_birth_cohort1830M1835 -0.56 0.12 -0.80 -0.31 600
## hu_male1 0.28 0.05 0.19 0.37 3000
## hu_maternalage.factor1420 0.32 0.23 -0.13 0.80 3000
## hu_maternalage.factor3550 0.21 0.07 0.07 0.35 2307
## hu_paternalage.mean 0.12 0.08 -0.03 0.28 1565
## hu_paternal_loss01 0.62 0.19 0.27 1.00 3000
## hu_paternal_loss15 0.54 0.13 0.28 0.80 1714
## hu_paternal_loss510 0.16 0.11 -0.06 0.38 1461
## hu_paternal_loss1015 0.15 0.11 -0.06 0.35 1437
## hu_paternal_loss1520 0.10 0.11 -0.11 0.30 1485
## hu_paternal_loss2025 0.16 0.10 -0.04 0.35 1330
## hu_paternal_loss2530 0.04 0.09 -0.13 0.22 1342
## hu_paternal_loss3035 -0.04 0.09 -0.22 0.14 1426
## hu_paternal_loss3540 -0.03 0.09 -0.20 0.15 1316
## hu_paternal_loss4045 0.13 0.10 -0.08 0.33 1798
## hu_maternal_loss01 1.62 0.19 1.26 2.01 3000
## hu_maternal_loss15 0.59 0.12 0.35 0.84 1880
## hu_maternal_loss510 0.45 0.11 0.25 0.67 1748
## hu_maternal_loss1015 0.47 0.11 0.26 0.68 3000
## hu_maternal_loss1520 0.31 0.11 0.10 0.53 3000
## hu_maternal_loss2025 0.27 0.10 0.06 0.47 3000
## hu_maternal_loss2530 0.18 0.09 0.00 0.37 1682
## hu_maternal_loss3035 0.21 0.09 0.04 0.38 1636
## hu_maternal_loss3540 0.06 0.08 -0.10 0.22 3000
## hu_maternal_loss4045 0.27 0.08 0.11 0.44 3000
## hu_nr.siblings 0.03 0.01 0.00 0.05 2275
## hu_dependent_sibs_f5y 0.14 0.03 0.09 0.19 3000
## Rhat
## Intercept 1.00
## paternalage 1.00
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.00
## birth_cohort1770M1775 1.00
## birth_cohort1775M1780 1.00
## birth_cohort1780M1785 1.00
## birth_cohort1785M1790 1.00
## birth_cohort1790M1795 1.00
## birth_cohort1795M1800 1.00
## birth_cohort1800M1805 1.00
## birth_cohort1805M1810 1.01
## birth_cohort1810M1815 1.00
## birth_cohort1815M1820 1.00
## birth_cohort1820M1825 1.01
## birth_cohort1825M1830 1.00
## birth_cohort1830M1835 1.00
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.00
## paternal_loss01 1.00
## paternal_loss15 1.01
## paternal_loss510 1.01
## paternal_loss1015 1.01
## paternal_loss1520 1.01
## paternal_loss2025 1.01
## paternal_loss2530 1.01
## paternal_loss3035 1.01
## paternal_loss3540 1.01
## paternal_loss4045 1.01
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## nr.siblings 1.00
## dependent_sibs_f5y 1.00
## hu_Intercept 1.00
## hu_paternalage 1.00
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.01
## hu_birth_cohort1775M1780 1.00
## hu_birth_cohort1780M1785 1.00
## hu_birth_cohort1785M1790 1.00
## hu_birth_cohort1790M1795 1.00
## hu_birth_cohort1795M1800 1.00
## hu_birth_cohort1800M1805 1.00
## hu_birth_cohort1805M1810 1.00
## hu_birth_cohort1810M1815 1.00
## hu_birth_cohort1815M1820 1.00
## hu_birth_cohort1820M1825 1.00
## hu_birth_cohort1825M1830 1.00
## hu_birth_cohort1830M1835 1.00
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.00
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_nr.siblings 1.00
## hu_dependent_sibs_f5y 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.19 | 4.473 | 6.031 |
paternalage | 0.9838 | 0.9339 | 1.035 |
birth_cohort1760M1765 | 0.9977 | 0.88 | 1.124 |
birth_cohort1765M1770 | 0.8861 | 0.793 | 0.9916 |
birth_cohort1770M1775 | 0.8931 | 0.7988 | 1.004 |
birth_cohort1775M1780 | 0.9752 | 0.8723 | 1.091 |
birth_cohort1780M1785 | 0.8935 | 0.8001 | 1.001 |
birth_cohort1785M1790 | 0.9095 | 0.8181 | 1.012 |
birth_cohort1790M1795 | 0.9265 | 0.8346 | 1.03 |
birth_cohort1795M1800 | 0.9014 | 0.8205 | 0.9937 |
birth_cohort1800M1805 | 0.8943 | 0.8117 | 0.9829 |
birth_cohort1805M1810 | 0.8738 | 0.7935 | 0.9628 |
birth_cohort1810M1815 | 0.9063 | 0.8242 | 0.9953 |
birth_cohort1815M1820 | 0.8695 | 0.794 | 0.9498 |
birth_cohort1820M1825 | 0.8315 | 0.7574 | 0.9094 |
birth_cohort1825M1830 | 0.8097 | 0.7375 | 0.8868 |
birth_cohort1830M1835 | 0.8325 | 0.7573 | 0.9136 |
male1 | 1.083 | 1.048 | 1.12 |
maternalage.factor1420 | 0.935 | 0.7719 | 1.118 |
maternalage.factor3550 | 0.9985 | 0.9464 | 1.053 |
paternalage.mean | 1.014 | 0.958 | 1.079 |
paternal_loss01 | 0.8704 | 0.7521 | 1.006 |
paternal_loss15 | 0.9576 | 0.8654 | 1.06 |
paternal_loss510 | 0.9316 | 0.8572 | 1.015 |
paternal_loss1015 | 1.004 | 0.9291 | 1.088 |
paternal_loss1520 | 0.9076 | 0.8409 | 0.9802 |
paternal_loss2025 | 0.8862 | 0.8238 | 0.953 |
paternal_loss2530 | 0.989 | 0.9245 | 1.056 |
paternal_loss3035 | 0.9699 | 0.9084 | 1.034 |
paternal_loss3540 | 0.9869 | 0.9265 | 1.052 |
paternal_loss4045 | 0.9889 | 0.9189 | 1.062 |
maternal_loss01 | 1.11 | 0.9523 | 1.294 |
maternal_loss15 | 0.9744 | 0.8903 | 1.067 |
maternal_loss510 | 1.07 | 0.9889 | 1.154 |
maternal_loss1015 | 1.03 | 0.9508 | 1.118 |
maternal_loss1520 | 1.006 | 0.9281 | 1.092 |
maternal_loss2025 | 1.008 | 0.9355 | 1.087 |
maternal_loss2530 | 0.9775 | 0.9153 | 1.045 |
maternal_loss3035 | 0.9485 | 0.887 | 1.012 |
maternal_loss3540 | 0.9682 | 0.9131 | 1.025 |
maternal_loss4045 | 0.9717 | 0.9149 | 1.034 |
nr.siblings | 1.003 | 0.9932 | 1.013 |
dependent_sibs_f5y | 0.9959 | 0.9768 | 1.015 |
hu_Intercept | 0.6651 | 0.4483 | 0.9853 |
hu_paternalage | 0.9608 | 0.8337 | 1.103 |
hu_birth_cohort1760M1765 | 0.9701 | 0.704 | 1.346 |
hu_birth_cohort1765M1770 | 0.735 | 0.5487 | 0.9851 |
hu_birth_cohort1770M1775 | 0.9369 | 0.7024 | 1.247 |
hu_birth_cohort1775M1780 | 0.8299 | 0.6219 | 1.109 |
hu_birth_cohort1780M1785 | 0.7797 | 0.5791 | 1.055 |
hu_birth_cohort1785M1790 | 0.6622 | 0.496 | 0.8794 |
hu_birth_cohort1790M1795 | 0.7381 | 0.5651 | 0.9704 |
hu_birth_cohort1795M1800 | 0.6233 | 0.4862 | 0.7997 |
hu_birth_cohort1800M1805 | 0.5858 | 0.4538 | 0.7514 |
hu_birth_cohort1805M1810 | 0.7743 | 0.6003 | 0.9882 |
hu_birth_cohort1810M1815 | 0.6487 | 0.505 | 0.8205 |
hu_birth_cohort1815M1820 | 0.4932 | 0.3871 | 0.6221 |
hu_birth_cohort1820M1825 | 0.5814 | 0.4598 | 0.7426 |
hu_birth_cohort1825M1830 | 0.5826 | 0.4599 | 0.731 |
hu_birth_cohort1830M1835 | 0.5731 | 0.4482 | 0.7324 |
hu_male1 | 1.32 | 1.21 | 1.447 |
hu_maternalage.factor1420 | 1.376 | 0.8783 | 2.22 |
hu_maternalage.factor3550 | 1.232 | 1.075 | 1.425 |
hu_paternalage.mean | 1.133 | 0.9695 | 1.33 |
hu_paternal_loss01 | 1.861 | 1.306 | 2.706 |
hu_paternal_loss15 | 1.713 | 1.323 | 2.222 |
hu_paternal_loss510 | 1.171 | 0.9378 | 1.467 |
hu_paternal_loss1015 | 1.162 | 0.9371 | 1.424 |
hu_paternal_loss1520 | 1.1 | 0.8922 | 1.346 |
hu_paternal_loss2025 | 1.17 | 0.9618 | 1.414 |
hu_paternal_loss2530 | 1.045 | 0.8747 | 1.251 |
hu_paternal_loss3035 | 0.9569 | 0.8026 | 1.147 |
hu_paternal_loss3540 | 0.9731 | 0.8178 | 1.165 |
hu_paternal_loss4045 | 1.134 | 0.926 | 1.386 |
hu_maternal_loss01 | 5.078 | 3.52 | 7.43 |
hu_maternal_loss15 | 1.799 | 1.424 | 2.307 |
hu_maternal_loss510 | 1.576 | 1.282 | 1.951 |
hu_maternal_loss1015 | 1.596 | 1.293 | 1.973 |
hu_maternal_loss1520 | 1.369 | 1.11 | 1.702 |
hu_maternal_loss2025 | 1.305 | 1.064 | 1.603 |
hu_maternal_loss2530 | 1.202 | 1.002 | 1.449 |
hu_maternal_loss3035 | 1.235 | 1.042 | 1.457 |
hu_maternal_loss3540 | 1.057 | 0.9018 | 1.248 |
hu_maternal_loss4045 | 1.31 | 1.111 | 1.547 |
hu_nr.siblings | 1.026 | 0.9996 | 1.053 |
hu_dependent_sibs_f5y | 1.153 | 1.098 | 1.212 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.21 | [1.87;2.57] | [1.99;2.45] |
estimate father 35y | 2.23 | [1.86;2.62] | [1.99;2.49] |
percentage change | 0.68 | [-8.48;10.93] | [-5.24;7.31] |
OR/IRR | 0.98 | [0.93;1.04] | [0.95;1.02] |
OR hurdle | 0.96 | [0.83;1.1] | [0.88;1.05] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r4_control_dependent_sibs.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
Plausibly, being first-born has a different effect, when one is an only child as opposed to having two siblings, etc. Here, we allow for such an interaction effect.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings * nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents) + older_siblings:nr.siblings
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.23 0.01 0.20 0.25 1165 1.00
## sd(hu_Intercept) 0.48 0.05 0.38 0.56 657 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI
## Intercept 1.68 0.08 1.52 1.83
## paternalage 0.06 0.06 -0.05 0.16
## birth_cohort1760M1765 0.00 0.06 -0.13 0.12
## birth_cohort1765M1770 -0.12 0.06 -0.23 -0.01
## birth_cohort1770M1775 -0.12 0.06 -0.23 0.00
## birth_cohort1775M1780 -0.03 0.05 -0.13 0.08
## birth_cohort1780M1785 -0.11 0.06 -0.22 0.00
## birth_cohort1785M1790 -0.10 0.05 -0.20 0.02
## birth_cohort1790M1795 -0.08 0.05 -0.17 0.02
## birth_cohort1795M1800 -0.11 0.05 -0.20 -0.01
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.02
## birth_cohort1805M1810 -0.14 0.05 -0.23 -0.04
## birth_cohort1810M1815 -0.10 0.05 -0.19 0.00
## birth_cohort1815M1820 -0.14 0.04 -0.23 -0.06
## birth_cohort1820M1825 -0.19 0.04 -0.27 -0.10
## birth_cohort1825M1830 -0.21 0.04 -0.29 -0.12
## birth_cohort1830M1835 -0.18 0.05 -0.27 -0.09
## male1 0.08 0.02 0.04 0.11
## maternalage.factor1420 -0.04 0.09 -0.23 0.14
## maternalage.factor3550 0.00 0.03 -0.06 0.05
## paternalage.mean -0.06 0.06 -0.17 0.05
## paternal_loss01 -0.15 0.08 -0.31 -0.01
## paternal_loss15 -0.04 0.05 -0.14 0.06
## paternal_loss510 -0.07 0.04 -0.15 0.01
## paternal_loss1015 0.01 0.04 -0.07 0.08
## paternal_loss1520 -0.09 0.04 -0.17 -0.02
## paternal_loss2025 -0.12 0.04 -0.20 -0.05
## paternal_loss2530 -0.01 0.03 -0.07 0.06
## paternal_loss3035 -0.03 0.03 -0.09 0.03
## paternal_loss3540 -0.01 0.03 -0.07 0.05
## paternal_loss4045 -0.01 0.04 -0.08 0.06
## maternal_loss01 0.10 0.08 -0.05 0.25
## maternal_loss15 -0.02 0.05 -0.11 0.08
## maternal_loss510 0.07 0.04 -0.01 0.15
## maternal_loss1015 0.03 0.04 -0.06 0.10
## maternal_loss1520 0.00 0.04 -0.08 0.08
## maternal_loss2025 0.00 0.04 -0.07 0.08
## maternal_loss2530 -0.02 0.03 -0.09 0.05
## maternal_loss3035 -0.05 0.03 -0.11 0.01
## maternal_loss3540 -0.03 0.03 -0.09 0.02
## maternal_loss4045 -0.03 0.03 -0.09 0.04
## older_siblings1 0.00 0.05 -0.11 0.10
## older_siblings2 -0.09 0.07 -0.23 0.04
## older_siblings3 -0.10 0.09 -0.27 0.08
## older_siblings4 -0.06 0.13 -0.30 0.19
## older_siblings5P -0.18 0.11 -0.41 0.04
## nr.siblings 0.00 0.01 -0.02 0.02
## last_born1 -0.04 0.02 -0.09 0.00
## older_siblings1:nr.siblings 0.01 0.01 -0.01 0.03
## older_siblings2:nr.siblings 0.01 0.01 -0.01 0.03
## older_siblings3:nr.siblings 0.01 0.01 -0.02 0.04
## older_siblings4:nr.siblings 0.00 0.02 -0.04 0.04
## older_siblings5P:nr.siblings 0.02 0.01 -0.01 0.04
## hu_Intercept -0.26 0.21 -0.66 0.15
## hu_paternalage 0.29 0.15 -0.02 0.60
## hu_birth_cohort1760M1765 -0.06 0.17 -0.38 0.27
## hu_birth_cohort1765M1770 -0.32 0.15 -0.61 -0.03
## hu_birth_cohort1770M1775 -0.07 0.14 -0.35 0.21
## hu_birth_cohort1775M1780 -0.20 0.15 -0.49 0.09
## hu_birth_cohort1780M1785 -0.28 0.15 -0.57 0.00
## hu_birth_cohort1785M1790 -0.43 0.14 -0.71 -0.15
## hu_birth_cohort1790M1795 -0.32 0.14 -0.59 -0.06
## hu_birth_cohort1795M1800 -0.47 0.13 -0.73 -0.22
## hu_birth_cohort1800M1805 -0.54 0.13 -0.79 -0.29
## hu_birth_cohort1805M1810 -0.27 0.12 -0.52 -0.03
## hu_birth_cohort1810M1815 -0.45 0.12 -0.69 -0.20
## hu_birth_cohort1815M1820 -0.71 0.12 -0.94 -0.48
## hu_birth_cohort1820M1825 -0.53 0.12 -0.76 -0.30
## hu_birth_cohort1825M1830 -0.55 0.12 -0.79 -0.32
## hu_birth_cohort1830M1835 -0.56 0.12 -0.80 -0.34
## hu_male1 0.27 0.05 0.18 0.36
## hu_maternalage.factor1420 0.28 0.24 -0.17 0.74
## hu_maternalage.factor3550 0.15 0.07 0.02 0.30
## hu_paternalage.mean -0.21 0.16 -0.51 0.10
## hu_paternal_loss01 0.59 0.18 0.24 0.94
## hu_paternal_loss15 0.53 0.13 0.28 0.79
## hu_paternal_loss510 0.19 0.11 -0.03 0.41
## hu_paternal_loss1015 0.15 0.10 -0.05 0.35
## hu_paternal_loss1520 0.10 0.10 -0.10 0.29
## hu_paternal_loss2025 0.15 0.10 -0.04 0.34
## hu_paternal_loss2530 0.06 0.09 -0.13 0.24
## hu_paternal_loss3035 -0.03 0.09 -0.21 0.15
## hu_paternal_loss3540 -0.02 0.09 -0.20 0.17
## hu_paternal_loss4045 0.15 0.10 -0.05 0.33
## hu_maternal_loss01 1.59 0.19 1.22 1.98
## hu_maternal_loss15 0.59 0.12 0.36 0.82
## hu_maternal_loss510 0.48 0.11 0.27 0.68
## hu_maternal_loss1015 0.47 0.11 0.25 0.68
## hu_maternal_loss1520 0.31 0.11 0.10 0.53
## hu_maternal_loss2025 0.26 0.10 0.06 0.46
## hu_maternal_loss2530 0.19 0.10 0.00 0.38
## hu_maternal_loss3035 0.22 0.09 0.05 0.39
## hu_maternal_loss3540 0.06 0.08 -0.10 0.22
## hu_maternal_loss4045 0.28 0.09 0.11 0.45
## hu_older_siblings1 -0.08 0.14 -0.36 0.18
## hu_older_siblings2 -0.25 0.18 -0.61 0.08
## hu_older_siblings3 -0.43 0.23 -0.89 0.02
## hu_older_siblings4 -0.70 0.32 -1.33 -0.07
## hu_older_siblings5P -0.63 0.30 -1.24 -0.06
## hu_nr.siblings 0.09 0.02 0.04 0.14
## hu_last_born1 0.09 0.06 -0.03 0.21
## hu_older_siblings1:nr.siblings 0.02 0.03 -0.04 0.07
## hu_older_siblings2:nr.siblings 0.02 0.03 -0.04 0.08
## hu_older_siblings3:nr.siblings 0.05 0.04 -0.03 0.12
## hu_older_siblings4:nr.siblings 0.08 0.05 -0.01 0.17
## hu_older_siblings5P:nr.siblings 0.01 0.04 -0.06 0.08
## Eff.Sample Rhat
## Intercept 1301 1.01
## paternalage 1058 1.00
## birth_cohort1760M1765 1434 1.01
## birth_cohort1765M1770 834 1.01
## birth_cohort1770M1775 857 1.01
## birth_cohort1775M1780 792 1.02
## birth_cohort1780M1785 764 1.01
## birth_cohort1785M1790 931 1.01
## birth_cohort1790M1795 740 1.02
## birth_cohort1795M1800 707 1.02
## birth_cohort1800M1805 376 1.02
## birth_cohort1805M1810 636 1.02
## birth_cohort1810M1815 555 1.02
## birth_cohort1815M1820 520 1.02
## birth_cohort1820M1825 582 1.02
## birth_cohort1825M1830 299 1.02
## birth_cohort1830M1835 350 1.02
## male1 3000 1.00
## maternalage.factor1420 3000 1.00
## maternalage.factor3550 2380 1.00
## paternalage.mean 1094 1.00
## paternal_loss01 3000 1.00
## paternal_loss15 1394 1.00
## paternal_loss510 1468 1.00
## paternal_loss1015 1236 1.00
## paternal_loss1520 1114 1.00
## paternal_loss2025 1078 1.00
## paternal_loss2530 1025 1.00
## paternal_loss3035 1034 1.00
## paternal_loss3540 1359 1.00
## paternal_loss4045 3000 1.00
## maternal_loss01 3000 1.00
## maternal_loss15 2039 1.00
## maternal_loss510 3000 1.00
## maternal_loss1015 2324 1.00
## maternal_loss1520 2293 1.00
## maternal_loss2025 3000 1.00
## maternal_loss2530 1532 1.00
## maternal_loss3035 2002 1.00
## maternal_loss3540 2124 1.00
## maternal_loss4045 3000 1.00
## older_siblings1 1379 1.00
## older_siblings2 1277 1.00
## older_siblings3 1375 1.00
## older_siblings4 1116 1.00
## older_siblings5P 1329 1.00
## nr.siblings 1732 1.00
## last_born1 3000 1.00
## older_siblings1:nr.siblings 1592 1.00
## older_siblings2:nr.siblings 1533 1.00
## older_siblings3:nr.siblings 1850 1.00
## older_siblings4:nr.siblings 1549 1.00
## older_siblings5P:nr.siblings 1474 1.00
## hu_Intercept 1102 1.00
## hu_paternalage 986 1.00
## hu_birth_cohort1760M1765 3000 1.00
## hu_birth_cohort1765M1770 1173 1.01
## hu_birth_cohort1770M1775 1084 1.00
## hu_birth_cohort1775M1780 1050 1.01
## hu_birth_cohort1780M1785 1003 1.00
## hu_birth_cohort1785M1790 964 1.00
## hu_birth_cohort1790M1795 879 1.01
## hu_birth_cohort1795M1800 850 1.01
## hu_birth_cohort1800M1805 850 1.01
## hu_birth_cohort1805M1810 824 1.01
## hu_birth_cohort1810M1815 800 1.01
## hu_birth_cohort1815M1820 752 1.01
## hu_birth_cohort1820M1825 788 1.01
## hu_birth_cohort1825M1830 761 1.01
## hu_birth_cohort1830M1835 773 1.01
## hu_male1 3000 1.00
## hu_maternalage.factor1420 3000 1.00
## hu_maternalage.factor3550 3000 1.00
## hu_paternalage.mean 1009 1.00
## hu_paternal_loss01 3000 1.00
## hu_paternal_loss15 1531 1.00
## hu_paternal_loss510 1394 1.00
## hu_paternal_loss1015 1294 1.01
## hu_paternal_loss1520 1084 1.01
## hu_paternal_loss2025 1194 1.00
## hu_paternal_loss2530 1128 1.01
## hu_paternal_loss3035 1163 1.00
## hu_paternal_loss3540 1342 1.00
## hu_paternal_loss4045 1459 1.00
## hu_maternal_loss01 3000 1.00
## hu_maternal_loss15 1714 1.00
## hu_maternal_loss510 1569 1.00
## hu_maternal_loss1015 3000 1.00
## hu_maternal_loss1520 3000 1.00
## hu_maternal_loss2025 1688 1.00
## hu_maternal_loss2530 1578 1.00
## hu_maternal_loss3035 1549 1.00
## hu_maternal_loss3540 1662 1.00
## hu_maternal_loss4045 1586 1.00
## hu_older_siblings1 1127 1.00
## hu_older_siblings2 1045 1.00
## hu_older_siblings3 959 1.00
## hu_older_siblings4 1213 1.00
## hu_older_siblings5P 1182 1.00
## hu_nr.siblings 1614 1.00
## hu_last_born1 3000 1.00
## hu_older_siblings1:nr.siblings 1334 1.00
## hu_older_siblings2:nr.siblings 1540 1.00
## hu_older_siblings3:nr.siblings 1409 1.00
## hu_older_siblings4:nr.siblings 1792 1.00
## hu_older_siblings5P:nr.siblings 1831 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.35 | 4.576 | 6.245 |
paternalage | 1.059 | 0.9503 | 1.178 |
birth_cohort1760M1765 | 0.9967 | 0.8816 | 1.127 |
birth_cohort1765M1770 | 0.8892 | 0.7966 | 0.9942 |
birth_cohort1770M1775 | 0.8893 | 0.7932 | 1.002 |
birth_cohort1775M1780 | 0.9739 | 0.8743 | 1.084 |
birth_cohort1780M1785 | 0.8949 | 0.8007 | 1.003 |
birth_cohort1785M1790 | 0.9092 | 0.8172 | 1.015 |
birth_cohort1790M1795 | 0.9266 | 0.8415 | 1.02 |
birth_cohort1795M1800 | 0.8972 | 0.8175 | 0.9857 |
birth_cohort1800M1805 | 0.8921 | 0.8131 | 0.9791 |
birth_cohort1805M1810 | 0.8711 | 0.793 | 0.9574 |
birth_cohort1810M1815 | 0.9065 | 0.8278 | 0.9977 |
birth_cohort1815M1820 | 0.867 | 0.7958 | 0.9452 |
birth_cohort1820M1825 | 0.8301 | 0.7607 | 0.9049 |
birth_cohort1825M1830 | 0.8101 | 0.7454 | 0.8852 |
birth_cohort1830M1835 | 0.8328 | 0.7629 | 0.9139 |
male1 | 1.08 | 1.046 | 1.116 |
maternalage.factor1420 | 0.9595 | 0.7944 | 1.149 |
maternalage.factor3550 | 1 | 0.9447 | 1.055 |
paternalage.mean | 0.9414 | 0.8402 | 1.05 |
paternal_loss01 | 0.859 | 0.7351 | 0.99 |
paternal_loss15 | 0.9607 | 0.8696 | 1.057 |
paternal_loss510 | 0.9322 | 0.8582 | 1.013 |
paternal_loss1015 | 1.007 | 0.9315 | 1.088 |
paternal_loss1520 | 0.9097 | 0.8434 | 0.9804 |
paternal_loss2025 | 0.8856 | 0.8215 | 0.9527 |
paternal_loss2530 | 0.9906 | 0.9284 | 1.059 |
paternal_loss3035 | 0.9712 | 0.9103 | 1.034 |
paternal_loss3540 | 0.9908 | 0.9283 | 1.055 |
paternal_loss4045 | 0.9897 | 0.9196 | 1.061 |
maternal_loss01 | 1.104 | 0.9467 | 1.286 |
maternal_loss15 | 0.983 | 0.8964 | 1.081 |
maternal_loss510 | 1.074 | 0.9908 | 1.162 |
maternal_loss1015 | 1.026 | 0.9432 | 1.11 |
maternal_loss1520 | 1.003 | 0.9261 | 1.085 |
maternal_loss2025 | 1.004 | 0.9278 | 1.086 |
maternal_loss2530 | 0.979 | 0.9163 | 1.046 |
maternal_loss3035 | 0.9504 | 0.892 | 1.012 |
maternal_loss3540 | 0.9674 | 0.9137 | 1.025 |
maternal_loss4045 | 0.972 | 0.9121 | 1.037 |
older_siblings1 | 0.998 | 0.8987 | 1.104 |
older_siblings2 | 0.9139 | 0.7977 | 1.044 |
older_siblings3 | 0.9052 | 0.7663 | 1.08 |
older_siblings4 | 0.9425 | 0.7424 | 1.207 |
older_siblings5P | 0.8337 | 0.6631 | 1.043 |
nr.siblings | 1.003 | 0.9847 | 1.02 |
last_born1 | 0.9574 | 0.9157 | 1.002 |
older_siblings1:nr.siblings | 1.008 | 0.9869 | 1.029 |
older_siblings2:nr.siblings | 1.011 | 0.9883 | 1.036 |
older_siblings3:nr.siblings | 1.008 | 0.9791 | 1.036 |
older_siblings4:nr.siblings | 0.9982 | 0.9612 | 1.036 |
older_siblings5P:nr.siblings | 1.016 | 0.9882 | 1.045 |
hu_Intercept | 0.7733 | 0.5161 | 1.158 |
hu_paternalage | 1.337 | 0.9824 | 1.815 |
hu_birth_cohort1760M1765 | 0.939 | 0.6841 | 1.314 |
hu_birth_cohort1765M1770 | 0.7256 | 0.5412 | 0.9712 |
hu_birth_cohort1770M1775 | 0.9341 | 0.703 | 1.228 |
hu_birth_cohort1775M1780 | 0.8173 | 0.6149 | 1.09 |
hu_birth_cohort1780M1785 | 0.7537 | 0.5639 | 1.005 |
hu_birth_cohort1785M1790 | 0.6515 | 0.4932 | 0.8648 |
hu_birth_cohort1790M1795 | 0.7254 | 0.5564 | 0.9386 |
hu_birth_cohort1795M1800 | 0.6239 | 0.4809 | 0.8016 |
hu_birth_cohort1800M1805 | 0.5816 | 0.4528 | 0.7479 |
hu_birth_cohort1805M1810 | 0.7605 | 0.5953 | 0.9682 |
hu_birth_cohort1810M1815 | 0.6391 | 0.4998 | 0.8168 |
hu_birth_cohort1815M1820 | 0.4919 | 0.3909 | 0.6165 |
hu_birth_cohort1820M1825 | 0.5895 | 0.466 | 0.741 |
hu_birth_cohort1825M1830 | 0.5764 | 0.4546 | 0.7271 |
hu_birth_cohort1830M1835 | 0.5702 | 0.4481 | 0.7148 |
hu_male1 | 1.315 | 1.203 | 1.435 |
hu_maternalage.factor1420 | 1.317 | 0.8415 | 2.091 |
hu_maternalage.factor3550 | 1.165 | 1.015 | 1.353 |
hu_paternalage.mean | 0.8104 | 0.5975 | 1.108 |
hu_paternal_loss01 | 1.802 | 1.267 | 2.563 |
hu_paternal_loss15 | 1.707 | 1.326 | 2.214 |
hu_paternal_loss510 | 1.211 | 0.9738 | 1.504 |
hu_paternal_loss1015 | 1.168 | 0.9514 | 1.423 |
hu_paternal_loss1520 | 1.107 | 0.908 | 1.333 |
hu_paternal_loss2025 | 1.167 | 0.9628 | 1.404 |
hu_paternal_loss2530 | 1.063 | 0.8782 | 1.274 |
hu_paternal_loss3035 | 0.9751 | 0.8142 | 1.158 |
hu_paternal_loss3540 | 0.9834 | 0.8227 | 1.179 |
hu_paternal_loss4045 | 1.156 | 0.9528 | 1.397 |
hu_maternal_loss01 | 4.886 | 3.396 | 7.216 |
hu_maternal_loss15 | 1.799 | 1.43 | 2.268 |
hu_maternal_loss510 | 1.617 | 1.306 | 1.982 |
hu_maternal_loss1015 | 1.597 | 1.289 | 1.97 |
hu_maternal_loss1520 | 1.367 | 1.103 | 1.698 |
hu_maternal_loss2025 | 1.293 | 1.067 | 1.589 |
hu_maternal_loss2530 | 1.207 | 0.9979 | 1.468 |
hu_maternal_loss3035 | 1.24 | 1.052 | 1.475 |
hu_maternal_loss3540 | 1.064 | 0.9069 | 1.244 |
hu_maternal_loss4045 | 1.32 | 1.116 | 1.561 |
hu_older_siblings1 | 0.9231 | 0.699 | 1.201 |
hu_older_siblings2 | 0.7773 | 0.545 | 1.085 |
hu_older_siblings3 | 0.648 | 0.4089 | 1.021 |
hu_older_siblings4 | 0.4962 | 0.2651 | 0.935 |
hu_older_siblings5P | 0.5307 | 0.2898 | 0.9454 |
hu_nr.siblings | 1.096 | 1.046 | 1.147 |
hu_last_born1 | 1.092 | 0.9661 | 1.236 |
hu_older_siblings1:nr.siblings | 1.017 | 0.9636 | 1.075 |
hu_older_siblings2:nr.siblings | 1.021 | 0.9612 | 1.087 |
hu_older_siblings3:nr.siblings | 1.046 | 0.9746 | 1.123 |
hu_older_siblings4:nr.siblings | 1.082 | 0.9879 | 1.188 |
hu_older_siblings5P:nr.siblings | 1.009 | 0.9388 | 1.086 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.3 | [1.95;2.72] | [2.07;2.56] |
estimate father 35y | 2.05 | [1.63;2.58] | [1.75;2.39] |
percentage change | -10.72 | [-28.05;9.92] | [-22.69;2.31] |
OR/IRR | 1.06 | [0.95;1.18] | [0.99;1.14] |
OR hurdle | 1.34 | [0.98;1.81] | [1.09;1.63] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r5_birth_order_interact_siblings.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
Paternal age and birth order are highly collinear with each other and with maternal age. Therefore, the choice to include this predictor widens standard errors for each predictor and may be disputed. Here we show what happens when we simply omit the birth order control.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + nr.siblings + (1 | idParents)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + nr.siblings + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.23 0.01 0.20 0.25 909 1.00
## sd(hu_Intercept) 0.47 0.05 0.38 0.56 576 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.64 0.08 1.49 1.79 1280
## paternalage -0.02 0.03 -0.07 0.03 1190
## birth_cohort1760M1765 0.00 0.06 -0.12 0.12 1518
## birth_cohort1765M1770 -0.12 0.06 -0.23 -0.01 1130
## birth_cohort1770M1775 -0.11 0.06 -0.22 -0.01 991
## birth_cohort1775M1780 -0.02 0.05 -0.13 0.08 954
## birth_cohort1780M1785 -0.11 0.06 -0.22 0.00 1093
## birth_cohort1785M1790 -0.09 0.05 -0.20 0.02 933
## birth_cohort1790M1795 -0.07 0.05 -0.18 0.02 836
## birth_cohort1795M1800 -0.10 0.05 -0.20 -0.01 823
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.02 811
## birth_cohort1805M1810 -0.13 0.05 -0.23 -0.04 822
## birth_cohort1810M1815 -0.10 0.05 -0.19 -0.01 759
## birth_cohort1815M1820 -0.14 0.04 -0.23 -0.05 734
## birth_cohort1820M1825 -0.18 0.04 -0.27 -0.10 715
## birth_cohort1825M1830 -0.21 0.04 -0.30 -0.12 749
## birth_cohort1830M1835 -0.18 0.05 -0.28 -0.09 778
## male1 0.08 0.02 0.05 0.11 3000
## maternalage.factor1420 -0.06 0.09 -0.24 0.10 3000
## maternalage.factor3550 0.00 0.03 -0.05 0.05 2155
## paternalage.mean 0.02 0.03 -0.05 0.08 1426
## paternal_loss01 -0.14 0.07 -0.29 0.02 3000
## paternal_loss15 -0.04 0.05 -0.14 0.06 1846
## paternal_loss510 -0.07 0.04 -0.15 0.01 1489
## paternal_loss1015 0.01 0.04 -0.07 0.09 1320
## paternal_loss1520 -0.10 0.04 -0.17 -0.02 1435
## paternal_loss2025 -0.12 0.04 -0.19 -0.05 1391
## paternal_loss2530 -0.01 0.03 -0.08 0.06 1240
## paternal_loss3035 -0.03 0.03 -0.09 0.04 1285
## paternal_loss3540 -0.01 0.03 -0.07 0.05 1391
## paternal_loss4045 -0.01 0.04 -0.08 0.06 1738
## maternal_loss01 0.11 0.08 -0.05 0.26 3000
## maternal_loss15 -0.02 0.05 -0.12 0.07 1901
## maternal_loss510 0.07 0.04 -0.01 0.15 3000
## maternal_loss1015 0.03 0.04 -0.05 0.11 1839
## maternal_loss1520 0.01 0.04 -0.07 0.09 1825
## maternal_loss2025 0.01 0.04 -0.06 0.08 1795
## maternal_loss2530 -0.02 0.03 -0.09 0.04 1662
## maternal_loss3035 -0.05 0.03 -0.12 0.01 1858
## maternal_loss3540 -0.03 0.03 -0.09 0.03 1903
## maternal_loss4045 -0.03 0.03 -0.09 0.03 2055
## nr.siblings 0.00 0.00 -0.01 0.01 2552
## hu_Intercept -0.25 0.19 -0.63 0.13 703
## hu_paternalage -0.05 0.07 -0.18 0.09 1460
## hu_birth_cohort1760M1765 -0.05 0.17 -0.39 0.28 3000
## hu_birth_cohort1765M1770 -0.32 0.15 -0.60 -0.02 862
## hu_birth_cohort1770M1775 -0.08 0.15 -0.37 0.21 731
## hu_birth_cohort1775M1780 -0.20 0.15 -0.49 0.08 712
## hu_birth_cohort1780M1785 -0.28 0.15 -0.57 0.01 816
## hu_birth_cohort1785M1790 -0.43 0.14 -0.71 -0.14 757
## hu_birth_cohort1790M1795 -0.31 0.14 -0.59 -0.04 698
## hu_birth_cohort1795M1800 -0.47 0.13 -0.72 -0.21 516
## hu_birth_cohort1800M1805 -0.55 0.13 -0.81 -0.30 555
## hu_birth_cohort1805M1810 -0.28 0.13 -0.52 -0.03 546
## hu_birth_cohort1810M1815 -0.44 0.12 -0.68 -0.20 518
## hu_birth_cohort1815M1820 -0.71 0.12 -0.94 -0.49 444
## hu_birth_cohort1820M1825 -0.54 0.12 -0.77 -0.30 517
## hu_birth_cohort1825M1830 -0.56 0.12 -0.78 -0.32 503
## hu_birth_cohort1830M1835 -0.56 0.12 -0.80 -0.32 504
## hu_male1 0.27 0.04 0.19 0.36 3000
## hu_maternalage.factor1420 0.23 0.23 -0.23 0.69 3000
## hu_maternalage.factor3550 0.13 0.07 -0.01 0.27 2206
## hu_paternalage.mean 0.12 0.08 -0.03 0.28 1434
## hu_paternal_loss01 0.56 0.18 0.21 0.92 3000
## hu_paternal_loss15 0.54 0.13 0.30 0.79 1696
## hu_paternal_loss510 0.19 0.11 -0.03 0.42 1318
## hu_paternal_loss1015 0.15 0.11 -0.06 0.37 1589
## hu_paternal_loss1520 0.10 0.10 -0.09 0.30 1124
## hu_paternal_loss2025 0.15 0.10 -0.04 0.35 1197
## hu_paternal_loss2530 0.05 0.09 -0.13 0.24 975
## hu_paternal_loss3035 -0.03 0.09 -0.20 0.15 1216
## hu_paternal_loss3540 -0.02 0.09 -0.19 0.16 1342
## hu_paternal_loss4045 0.14 0.10 -0.06 0.33 3000
## hu_maternal_loss01 1.55 0.19 1.19 1.92 3000
## hu_maternal_loss15 0.60 0.12 0.36 0.83 3000
## hu_maternal_loss510 0.49 0.10 0.28 0.69 1905
## hu_maternal_loss1015 0.47 0.11 0.27 0.69 3000
## hu_maternal_loss1520 0.31 0.11 0.10 0.52 3000
## hu_maternal_loss2025 0.25 0.10 0.05 0.46 3000
## hu_maternal_loss2530 0.19 0.09 0.01 0.36 1609
## hu_maternal_loss3035 0.21 0.09 0.04 0.39 1886
## hu_maternal_loss3540 0.06 0.08 -0.09 0.21 3000
## hu_maternal_loss4045 0.27 0.09 0.10 0.44 3000
## hu_nr.siblings 0.06 0.01 0.04 0.08 3000
## Rhat
## Intercept 1.00
## paternalage 1.00
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.01
## birth_cohort1770M1775 1.00
## birth_cohort1775M1780 1.01
## birth_cohort1780M1785 1.00
## birth_cohort1785M1790 1.00
## birth_cohort1790M1795 1.01
## birth_cohort1795M1800 1.01
## birth_cohort1800M1805 1.01
## birth_cohort1805M1810 1.01
## birth_cohort1810M1815 1.01
## birth_cohort1815M1820 1.01
## birth_cohort1820M1825 1.01
## birth_cohort1825M1830 1.01
## birth_cohort1830M1835 1.00
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.00
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.01
## paternal_loss3035 1.01
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## nr.siblings 1.00
## hu_Intercept 1.00
## hu_paternalage 1.00
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.01
## hu_birth_cohort1775M1780 1.01
## hu_birth_cohort1780M1785 1.00
## hu_birth_cohort1785M1790 1.01
## hu_birth_cohort1790M1795 1.01
## hu_birth_cohort1795M1800 1.01
## hu_birth_cohort1800M1805 1.01
## hu_birth_cohort1805M1810 1.01
## hu_birth_cohort1810M1815 1.01
## hu_birth_cohort1815M1820 1.01
## hu_birth_cohort1820M1825 1.01
## hu_birth_cohort1825M1830 1.01
## hu_birth_cohort1830M1835 1.01
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.00
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_nr.siblings 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.15 | 4.432 | 5.992 |
paternalage | 0.9814 | 0.9327 | 1.034 |
birth_cohort1760M1765 | 1.001 | 0.8846 | 1.128 |
birth_cohort1765M1770 | 0.8887 | 0.7925 | 0.991 |
birth_cohort1770M1775 | 0.8946 | 0.802 | 0.9949 |
birth_cohort1775M1780 | 0.9774 | 0.8794 | 1.086 |
birth_cohort1780M1785 | 0.8953 | 0.8012 | 1.001 |
birth_cohort1785M1790 | 0.9116 | 0.8182 | 1.015 |
birth_cohort1790M1795 | 0.9278 | 0.8384 | 1.023 |
birth_cohort1795M1800 | 0.9029 | 0.8179 | 0.9872 |
birth_cohort1800M1805 | 0.8959 | 0.8124 | 0.9793 |
birth_cohort1805M1810 | 0.877 | 0.7953 | 0.9622 |
birth_cohort1810M1815 | 0.9082 | 0.8289 | 0.9904 |
birth_cohort1815M1820 | 0.8707 | 0.7968 | 0.949 |
birth_cohort1820M1825 | 0.8329 | 0.7602 | 0.9063 |
birth_cohort1825M1830 | 0.8125 | 0.7433 | 0.8844 |
birth_cohort1830M1835 | 0.8341 | 0.7589 | 0.9113 |
male1 | 1.083 | 1.047 | 1.119 |
maternalage.factor1420 | 0.9381 | 0.7898 | 1.104 |
maternalage.factor3550 | 1.002 | 0.9518 | 1.055 |
paternalage.mean | 1.016 | 0.953 | 1.078 |
paternal_loss01 | 0.8736 | 0.7501 | 1.016 |
paternal_loss15 | 0.959 | 0.868 | 1.063 |
paternal_loss510 | 0.9336 | 0.8587 | 1.014 |
paternal_loss1015 | 1.006 | 0.9332 | 1.09 |
paternal_loss1520 | 0.9086 | 0.8413 | 0.9819 |
paternal_loss2025 | 0.8879 | 0.8247 | 0.9552 |
paternal_loss2530 | 0.9904 | 0.9241 | 1.06 |
paternal_loss3035 | 0.9709 | 0.9095 | 1.038 |
paternal_loss3540 | 0.9887 | 0.9287 | 1.05 |
paternal_loss4045 | 0.9907 | 0.9217 | 1.066 |
maternal_loss01 | 1.112 | 0.9553 | 1.294 |
maternal_loss15 | 0.976 | 0.8863 | 1.067 |
maternal_loss510 | 1.07 | 0.9875 | 1.158 |
maternal_loss1015 | 1.031 | 0.9484 | 1.121 |
maternal_loss1520 | 1.008 | 0.9293 | 1.09 |
maternal_loss2025 | 1.01 | 0.9384 | 1.088 |
maternal_loss2530 | 0.9784 | 0.9128 | 1.045 |
maternal_loss3035 | 0.9488 | 0.8898 | 1.01 |
maternal_loss3540 | 0.9688 | 0.9165 | 1.027 |
maternal_loss4045 | 0.9718 | 0.9171 | 1.032 |
nr.siblings | 1.002 | 0.994 | 1.01 |
hu_Intercept | 0.7815 | 0.5332 | 1.135 |
hu_paternalage | 0.9537 | 0.8335 | 1.096 |
hu_birth_cohort1760M1765 | 0.9486 | 0.6772 | 1.32 |
hu_birth_cohort1765M1770 | 0.7281 | 0.5502 | 0.9806 |
hu_birth_cohort1770M1775 | 0.9235 | 0.6914 | 1.233 |
hu_birth_cohort1775M1780 | 0.8153 | 0.6154 | 1.079 |
hu_birth_cohort1780M1785 | 0.7557 | 0.5675 | 1.009 |
hu_birth_cohort1785M1790 | 0.653 | 0.4926 | 0.8661 |
hu_birth_cohort1790M1795 | 0.7304 | 0.5517 | 0.9593 |
hu_birth_cohort1795M1800 | 0.6248 | 0.4851 | 0.8072 |
hu_birth_cohort1800M1805 | 0.577 | 0.4466 | 0.7388 |
hu_birth_cohort1805M1810 | 0.7564 | 0.5928 | 0.9689 |
hu_birth_cohort1810M1815 | 0.642 | 0.506 | 0.8165 |
hu_birth_cohort1815M1820 | 0.4911 | 0.3898 | 0.6156 |
hu_birth_cohort1820M1825 | 0.5854 | 0.4643 | 0.7421 |
hu_birth_cohort1825M1830 | 0.5739 | 0.4573 | 0.7268 |
hu_birth_cohort1830M1835 | 0.5719 | 0.4514 | 0.7231 |
hu_male1 | 1.314 | 1.204 | 1.436 |
hu_maternalage.factor1420 | 1.255 | 0.7912 | 1.992 |
hu_maternalage.factor3550 | 1.138 | 0.9915 | 1.305 |
hu_paternalage.mean | 1.131 | 0.966 | 1.323 |
hu_paternal_loss01 | 1.743 | 1.233 | 2.52 |
hu_paternal_loss15 | 1.711 | 1.346 | 2.206 |
hu_paternal_loss510 | 1.21 | 0.974 | 1.517 |
hu_paternal_loss1015 | 1.166 | 0.9452 | 1.446 |
hu_paternal_loss1520 | 1.105 | 0.9117 | 1.344 |
hu_paternal_loss2025 | 1.165 | 0.9634 | 1.422 |
hu_paternal_loss2530 | 1.053 | 0.8743 | 1.269 |
hu_paternal_loss3035 | 0.9727 | 0.8158 | 1.158 |
hu_paternal_loss3540 | 0.983 | 0.8245 | 1.172 |
hu_paternal_loss4045 | 1.148 | 0.9441 | 1.397 |
hu_maternal_loss01 | 4.713 | 3.282 | 6.831 |
hu_maternal_loss15 | 1.817 | 1.432 | 2.296 |
hu_maternal_loss510 | 1.624 | 1.323 | 1.984 |
hu_maternal_loss1015 | 1.606 | 1.304 | 2.002 |
hu_maternal_loss1520 | 1.36 | 1.101 | 1.68 |
hu_maternal_loss2025 | 1.289 | 1.054 | 1.576 |
hu_maternal_loss2530 | 1.205 | 1.006 | 1.439 |
hu_maternal_loss3035 | 1.236 | 1.042 | 1.47 |
hu_maternal_loss3540 | 1.064 | 0.9124 | 1.24 |
hu_maternal_loss4045 | 1.312 | 1.106 | 1.552 |
hu_nr.siblings | 1.062 | 1.04 | 1.085 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.17 | [1.83;2.53] | [1.95;2.4] |
estimate father 35y | 2.19 | [1.83;2.58] | [1.95;2.44] |
percentage change | 1.05 | [-8.48;11.09] | [-5.31;7.43] |
OR/IRR | 0.98 | [0.93;1.03] | [0.95;1.02] |
OR hurdle | 0.95 | [0.83;1.1] | [0.87;1.04] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r6_no_birth_order_control.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
We adjusted for parental loss very stringently, including covariates for parental loss up to age 45. Here we show what happens, when we only control for parental loss in the first, and the first five years of life.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 1000; warmup = 500; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.23 0.01 0.20 0.25 999 1.00
## sd(hu_Intercept) 0.48 0.04 0.39 0.56 816 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.64 0.09 1.46 1.82 1523
## paternalage 0.06 0.05 -0.04 0.16 645
## birth_cohort1760M1765 0.00 0.06 -0.12 0.12 910
## birth_cohort1765M1770 -0.11 0.06 -0.23 0.00 744
## birth_cohort1770M1775 -0.11 0.06 -0.22 0.00 696
## birth_cohort1775M1780 -0.02 0.05 -0.13 0.08 683
## birth_cohort1780M1785 -0.11 0.06 -0.22 0.00 643
## birth_cohort1785M1790 -0.10 0.05 -0.20 0.00 681
## birth_cohort1790M1795 -0.07 0.05 -0.18 0.03 604
## birth_cohort1795M1800 -0.11 0.05 -0.20 -0.01 517
## birth_cohort1800M1805 -0.11 0.05 -0.20 -0.01 552
## birth_cohort1805M1810 -0.13 0.05 -0.23 -0.04 520
## birth_cohort1810M1815 -0.10 0.05 -0.19 0.00 472
## birth_cohort1815M1820 -0.14 0.04 -0.22 -0.05 451
## birth_cohort1820M1825 -0.18 0.05 -0.27 -0.09 455
## birth_cohort1825M1830 -0.21 0.05 -0.30 -0.12 488
## birth_cohort1830M1835 -0.19 0.05 -0.28 -0.10 495
## male1 0.08 0.02 0.04 0.11 3000
## maternalage.factor1420 -0.04 0.09 -0.23 0.14 3000
## maternalage.factor3550 0.00 0.03 -0.06 0.05 3000
## paternalage.mean -0.07 0.05 -0.18 0.03 668
## paternal_loss01 -0.12 0.08 -0.28 0.04 3000
## paternal_losslater 0.00 0.04 -0.08 0.09 2622
## maternal_loss01 0.12 0.08 -0.04 0.28 3000
## maternal_losslater 0.02 0.04 -0.07 0.10 3000
## older_siblings1 0.03 0.03 -0.03 0.08 1243
## older_siblings2 -0.05 0.04 -0.12 0.02 731
## older_siblings3 -0.08 0.05 -0.17 0.01 692
## older_siblings4 -0.09 0.06 -0.21 0.03 697
## older_siblings5P -0.10 0.08 -0.25 0.05 612
## nr.siblings 0.01 0.01 0.00 0.02 757
## last_born1 -0.04 0.02 -0.09 0.01 3000
## hu_Intercept 0.76 0.22 0.33 1.20 1327
## hu_paternalage 0.40 0.14 0.11 0.69 639
## hu_birth_cohort1760M1765 -0.06 0.17 -0.39 0.27 3000
## hu_birth_cohort1765M1770 -0.29 0.15 -0.58 -0.01 820
## hu_birth_cohort1770M1775 -0.09 0.15 -0.37 0.20 832
## hu_birth_cohort1775M1780 -0.23 0.14 -0.50 0.06 781
## hu_birth_cohort1780M1785 -0.30 0.15 -0.59 -0.02 765
## hu_birth_cohort1785M1790 -0.44 0.14 -0.72 -0.17 775
## hu_birth_cohort1790M1795 -0.34 0.14 -0.61 -0.08 760
## hu_birth_cohort1795M1800 -0.48 0.13 -0.74 -0.24 651
## hu_birth_cohort1800M1805 -0.57 0.13 -0.81 -0.32 657
## hu_birth_cohort1805M1810 -0.32 0.12 -0.56 -0.07 626
## hu_birth_cohort1810M1815 -0.48 0.12 -0.72 -0.25 625
## hu_birth_cohort1815M1820 -0.73 0.12 -0.97 -0.51 609
## hu_birth_cohort1820M1825 -0.58 0.12 -0.81 -0.35 576
## hu_birth_cohort1825M1830 -0.58 0.12 -0.81 -0.36 571
## hu_birth_cohort1830M1835 -0.59 0.12 -0.83 -0.36 605
## hu_male1 0.27 0.04 0.19 0.36 3000
## hu_maternalage.factor1420 0.27 0.23 -0.18 0.75 3000
## hu_maternalage.factor3550 0.12 0.07 -0.03 0.26 3000
## hu_paternalage.mean -0.32 0.15 -0.61 -0.02 714
## hu_paternal_loss01 0.05 0.19 -0.32 0.42 3000
## hu_paternal_losslater -0.40 0.11 -0.62 -0.19 3000
## hu_maternal_loss01 0.99 0.20 0.60 1.37 3000
## hu_maternal_losslater -0.34 0.11 -0.55 -0.12 3000
## hu_older_siblings1 -0.02 0.07 -0.17 0.12 1230
## hu_older_siblings2 -0.18 0.10 -0.37 0.02 791
## hu_older_siblings3 -0.22 0.13 -0.47 0.02 560
## hu_older_siblings4 -0.27 0.16 -0.59 0.04 636
## hu_older_siblings5P -0.62 0.21 -1.03 -0.21 618
## hu_nr.siblings 0.10 0.02 0.06 0.13 864
## hu_last_born1 0.08 0.06 -0.03 0.21 3000
## Rhat
## Intercept 1.00
## paternalage 1.01
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.00
## birth_cohort1770M1775 1.01
## birth_cohort1775M1780 1.00
## birth_cohort1780M1785 1.00
## birth_cohort1785M1790 1.00
## birth_cohort1790M1795 1.00
## birth_cohort1795M1800 1.01
## birth_cohort1800M1805 1.01
## birth_cohort1805M1810 1.01
## birth_cohort1810M1815 1.01
## birth_cohort1815M1820 1.01
## birth_cohort1820M1825 1.01
## birth_cohort1825M1830 1.01
## birth_cohort1830M1835 1.01
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.01
## paternal_loss01 1.00
## paternal_losslater 1.00
## maternal_loss01 1.00
## maternal_losslater 1.00
## older_siblings1 1.00
## older_siblings2 1.01
## older_siblings3 1.01
## older_siblings4 1.01
## older_siblings5P 1.01
## nr.siblings 1.01
## last_born1 1.00
## hu_Intercept 1.00
## hu_paternalage 1.01
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.00
## hu_birth_cohort1775M1780 1.01
## hu_birth_cohort1780M1785 1.01
## hu_birth_cohort1785M1790 1.01
## hu_birth_cohort1790M1795 1.01
## hu_birth_cohort1795M1800 1.01
## hu_birth_cohort1800M1805 1.01
## hu_birth_cohort1805M1810 1.01
## hu_birth_cohort1810M1815 1.01
## hu_birth_cohort1815M1820 1.01
## hu_birth_cohort1820M1825 1.01
## hu_birth_cohort1825M1830 1.01
## hu_birth_cohort1830M1835 1.01
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.01
## hu_paternal_loss01 1.00
## hu_paternal_losslater 1.00
## hu_maternal_loss01 1.00
## hu_maternal_losslater 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.01
## hu_older_siblings3 1.01
## hu_older_siblings4 1.01
## hu_older_siblings5P 1.01
## hu_nr.siblings 1.01
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.159 | 4.297 | 6.16 |
paternalage | 1.06 | 0.9561 | 1.173 |
birth_cohort1760M1765 | 0.9992 | 0.8829 | 1.131 |
birth_cohort1765M1770 | 0.8922 | 0.796 | 0.9975 |
birth_cohort1770M1775 | 0.8975 | 0.8061 | 1.001 |
birth_cohort1775M1780 | 0.9763 | 0.8779 | 1.085 |
birth_cohort1780M1785 | 0.8965 | 0.8039 | 0.9984 |
birth_cohort1785M1790 | 0.904 | 0.8165 | 1.004 |
birth_cohort1790M1795 | 0.9283 | 0.8391 | 1.026 |
birth_cohort1795M1800 | 0.9 | 0.8176 | 0.9923 |
birth_cohort1800M1805 | 0.8969 | 0.8166 | 0.9855 |
birth_cohort1805M1810 | 0.8755 | 0.7953 | 0.9631 |
birth_cohort1810M1815 | 0.9087 | 0.8303 | 0.9951 |
birth_cohort1815M1820 | 0.8734 | 0.8007 | 0.951 |
birth_cohort1820M1825 | 0.8336 | 0.7635 | 0.9138 |
birth_cohort1825M1830 | 0.8068 | 0.7412 | 0.8847 |
birth_cohort1830M1835 | 0.8261 | 0.7549 | 0.9038 |
male1 | 1.08 | 1.044 | 1.116 |
maternalage.factor1420 | 0.9616 | 0.7961 | 1.154 |
maternalage.factor3550 | 0.9963 | 0.9414 | 1.054 |
paternalage.mean | 0.9291 | 0.8385 | 1.034 |
paternal_loss01 | 0.8881 | 0.7592 | 1.041 |
paternal_losslater | 1.001 | 0.9194 | 1.092 |
maternal_loss01 | 1.131 | 0.9583 | 1.329 |
maternal_losslater | 1.018 | 0.9365 | 1.107 |
older_siblings1 | 1.026 | 0.9723 | 1.081 |
older_siblings2 | 0.9539 | 0.8892 | 1.025 |
older_siblings3 | 0.9263 | 0.8466 | 1.014 |
older_siblings4 | 0.9104 | 0.8105 | 1.027 |
older_siblings5P | 0.9057 | 0.7779 | 1.055 |
nr.siblings | 1.01 | 0.9951 | 1.023 |
last_born1 | 0.9597 | 0.9151 | 1.006 |
hu_Intercept | 2.144 | 1.395 | 3.334 |
hu_paternalage | 1.489 | 1.122 | 1.992 |
hu_birth_cohort1760M1765 | 0.9411 | 0.6774 | 1.308 |
hu_birth_cohort1765M1770 | 0.7446 | 0.5578 | 0.995 |
hu_birth_cohort1770M1775 | 0.9176 | 0.6876 | 1.227 |
hu_birth_cohort1775M1780 | 0.7924 | 0.6045 | 1.057 |
hu_birth_cohort1780M1785 | 0.7383 | 0.5537 | 0.983 |
hu_birth_cohort1785M1790 | 0.6421 | 0.4856 | 0.8424 |
hu_birth_cohort1790M1795 | 0.7096 | 0.542 | 0.926 |
hu_birth_cohort1795M1800 | 0.6172 | 0.479 | 0.7886 |
hu_birth_cohort1800M1805 | 0.5662 | 0.4433 | 0.7283 |
hu_birth_cohort1805M1810 | 0.7255 | 0.5734 | 0.9308 |
hu_birth_cohort1810M1815 | 0.617 | 0.4891 | 0.7807 |
hu_birth_cohort1815M1820 | 0.4828 | 0.3805 | 0.6025 |
hu_birth_cohort1820M1825 | 0.5602 | 0.4453 | 0.7029 |
hu_birth_cohort1825M1830 | 0.5581 | 0.4456 | 0.7006 |
hu_birth_cohort1830M1835 | 0.5538 | 0.4353 | 0.6966 |
hu_male1 | 1.314 | 1.205 | 1.431 |
hu_maternalage.factor1420 | 1.307 | 0.8364 | 2.12 |
hu_maternalage.factor3550 | 1.125 | 0.9732 | 1.297 |
hu_paternalage.mean | 0.7291 | 0.5421 | 0.9831 |
hu_paternal_loss01 | 1.049 | 0.7244 | 1.52 |
hu_paternal_losslater | 0.6687 | 0.5376 | 0.83 |
hu_maternal_loss01 | 2.687 | 1.814 | 3.951 |
hu_maternal_losslater | 0.7149 | 0.5796 | 0.886 |
hu_older_siblings1 | 0.977 | 0.8456 | 1.133 |
hu_older_siblings2 | 0.8365 | 0.6903 | 1.016 |
hu_older_siblings3 | 0.8004 | 0.626 | 1.02 |
hu_older_siblings4 | 0.7613 | 0.5563 | 1.041 |
hu_older_siblings5P | 0.538 | 0.3579 | 0.8115 |
hu_nr.siblings | 1.104 | 1.065 | 1.144 |
hu_last_born1 | 1.087 | 0.9662 | 1.231 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 1.97 | [1.68;2.28] | [1.78;2.17] |
estimate father 35y | 1.61 | [1.29;1.97] | [1.39;1.84] |
percentage change | -18.19 | [-34.37;-0.24] | [-29.12;-6.7] |
OR/IRR | 1.06 | [0.96;1.17] | [0.99;1.13] |
OR hurdle | 1.49 | [1.12;1.99] | [1.24;1.79] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r7_less_parental_loss_control.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
Inheritance is linked to birth order and being male in several of the historical populations. Here, we adjust for the anchor being the first or last born adult son in a family. This implies that we control for our outcome to a certain extent, as “adult sons” cannot have died before adulthood, but a paternal age effect on mortality could still be detected for siblings other than the first- and last-born adults.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth.cohort + first_born_adult_male + last_born_adult_male + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + birth.cohort + first_born_adult_male + last_born_adult_male + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.23 0.01 0.20 0.25 1147 1.01
## sd(hu_Intercept) 0.54 0.04 0.45 0.63 900 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.68 0.08 1.52 1.83 1277
## paternalage 0.08 0.05 -0.02 0.19 771
## birth.cohort1760M1765 0.00 0.06 -0.13 0.13 3000
## birth.cohort1765M1770 -0.12 0.06 -0.23 -0.01 1023
## birth.cohort1770M1775 -0.12 0.06 -0.23 0.00 965
## birth.cohort1775M1780 -0.03 0.05 -0.13 0.08 827
## birth.cohort1780M1785 -0.11 0.06 -0.23 0.00 849
## birth.cohort1785M1790 -0.10 0.05 -0.20 0.01 806
## birth.cohort1790M1795 -0.08 0.05 -0.18 0.03 708
## birth.cohort1795M1800 -0.11 0.05 -0.20 -0.01 707
## birth.cohort1800M1805 -0.11 0.05 -0.21 -0.02 715
## birth.cohort1805M1810 -0.14 0.05 -0.23 -0.05 741
## birth.cohort1810M1815 -0.10 0.05 -0.19 0.00 657
## birth.cohort1815M1820 -0.14 0.04 -0.23 -0.05 582
## birth.cohort1820M1825 -0.19 0.05 -0.28 -0.09 640
## birth.cohort1825M1830 -0.21 0.05 -0.30 -0.12 657
## birth.cohort1830M1835 -0.18 0.05 -0.27 -0.09 644
## first_born_adult_male -0.03 0.03 -0.08 0.02 3000
## last_born_adult_male -0.05 0.03 -0.11 0.00 3000
## male1 0.12 0.03 0.07 0.17 2045
## maternalage.factor1420 -0.05 0.09 -0.23 0.12 3000
## maternalage.factor3550 0.00 0.03 -0.05 0.06 3000
## paternalage.mean -0.09 0.05 -0.20 0.01 721
## paternal_loss01 -0.15 0.08 -0.29 0.00 3000
## paternal_loss15 -0.03 0.05 -0.13 0.07 1929
## paternal_loss510 -0.07 0.04 -0.15 0.02 1671
## paternal_loss1015 0.01 0.04 -0.07 0.09 1542
## paternal_loss1520 -0.09 0.04 -0.17 -0.02 1442
## paternal_loss2025 -0.12 0.04 -0.20 -0.05 1555
## paternal_loss2530 -0.01 0.03 -0.08 0.06 1502
## paternal_loss3035 -0.03 0.03 -0.09 0.04 1496
## paternal_loss3540 -0.01 0.03 -0.07 0.05 1463
## paternal_loss4045 -0.01 0.04 -0.08 0.06 3000
## maternal_loss01 0.11 0.08 -0.05 0.26 3000
## maternal_loss15 -0.02 0.05 -0.12 0.07 3000
## maternal_loss510 0.07 0.04 -0.01 0.15 2160
## maternal_loss1015 0.03 0.04 -0.05 0.11 3000
## maternal_loss1520 0.00 0.04 -0.08 0.08 3000
## maternal_loss2025 0.00 0.04 -0.07 0.08 2035
## maternal_loss2530 -0.02 0.03 -0.09 0.04 1945
## maternal_loss3035 -0.05 0.03 -0.12 0.01 3000
## maternal_loss3540 -0.04 0.03 -0.09 0.02 3000
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## older_siblings1 0.02 0.03 -0.04 0.07 1740
## older_siblings2 -0.06 0.04 -0.13 0.01 1079
## older_siblings3 -0.09 0.05 -0.19 0.00 836
## older_siblings4 -0.11 0.06 -0.23 0.00 946
## older_siblings5P -0.12 0.08 -0.27 0.03 907
## nr.siblings 0.01 0.01 -0.01 0.02 1165
## last_born1 -0.05 0.02 -0.09 0.00 3000
## hu_Intercept 0.04 0.21 -0.37 0.44 1441
## hu_paternalage 0.39 0.15 0.10 0.68 770
## hu_birth.cohort1760M1765 -0.02 0.17 -0.35 0.33 3000
## hu_birth.cohort1765M1770 -0.31 0.15 -0.60 -0.02 1119
## hu_birth.cohort1770M1775 -0.10 0.15 -0.38 0.20 1087
## hu_birth.cohort1775M1780 -0.25 0.15 -0.53 0.05 958
## hu_birth.cohort1780M1785 -0.32 0.15 -0.63 -0.02 1078
## hu_birth.cohort1785M1790 -0.44 0.15 -0.72 -0.15 996
## hu_birth.cohort1790M1795 -0.33 0.14 -0.61 -0.06 896
## hu_birth.cohort1795M1800 -0.47 0.13 -0.71 -0.20 874
## hu_birth.cohort1800M1805 -0.54 0.13 -0.79 -0.28 738
## hu_birth.cohort1805M1810 -0.28 0.13 -0.53 -0.03 915
## hu_birth.cohort1810M1815 -0.43 0.12 -0.68 -0.19 827
## hu_birth.cohort1815M1820 -0.72 0.12 -0.95 -0.47 667
## hu_birth.cohort1820M1825 -0.54 0.12 -0.78 -0.31 639
## hu_birth.cohort1825M1830 -0.53 0.12 -0.77 -0.29 715
## hu_birth.cohort1830M1835 -0.57 0.12 -0.79 -0.31 692
## hu_first_born_adult_male -1.04 0.07 -1.18 -0.90 3000
## hu_last_born_adult_male -0.94 0.08 -1.09 -0.80 3000
## hu_male1 1.01 0.06 0.89 1.13 3000
## hu_maternalage.factor1420 0.25 0.23 -0.21 0.72 3000
## hu_maternalage.factor3550 0.17 0.08 0.01 0.31 3000
## hu_paternalage.mean -0.31 0.15 -0.60 -0.02 797
## hu_paternal_loss01 0.61 0.19 0.24 0.98 3000
## hu_paternal_loss15 0.57 0.13 0.31 0.83 1674
## hu_paternal_loss510 0.19 0.11 -0.03 0.41 1601
## hu_paternal_loss1015 0.15 0.11 -0.06 0.36 1605
## hu_paternal_loss1520 0.08 0.10 -0.12 0.28 1360
## hu_paternal_loss2025 0.12 0.10 -0.08 0.31 1325
## hu_paternal_loss2530 0.06 0.10 -0.13 0.25 1351
## hu_paternal_loss3035 -0.03 0.09 -0.22 0.15 1273
## hu_paternal_loss3540 -0.02 0.09 -0.19 0.16 1334
## hu_paternal_loss4045 0.11 0.10 -0.09 0.32 3000
## hu_maternal_loss01 1.58 0.19 1.22 1.96 3000
## hu_maternal_loss15 0.60 0.12 0.35 0.83 3000
## hu_maternal_loss510 0.49 0.11 0.28 0.70 3000
## hu_maternal_loss1015 0.49 0.11 0.27 0.71 3000
## hu_maternal_loss1520 0.30 0.11 0.09 0.52 3000
## hu_maternal_loss2025 0.24 0.11 0.03 0.46 3000
## hu_maternal_loss2530 0.18 0.10 -0.01 0.37 2159
## hu_maternal_loss3035 0.21 0.09 0.03 0.39 3000
## hu_maternal_loss3540 0.03 0.08 -0.13 0.19 3000
## hu_maternal_loss4045 0.29 0.09 0.12 0.48 3000
## hu_older_siblings1 -0.16 0.08 -0.31 -0.01 1298
## hu_older_siblings2 -0.37 0.10 -0.57 -0.17 902
## hu_older_siblings3 -0.46 0.13 -0.72 -0.20 791
## hu_older_siblings4 -0.49 0.16 -0.81 -0.17 745
## hu_older_siblings5P -0.85 0.21 -1.25 -0.43 716
## hu_nr.siblings 0.07 0.02 0.03 0.11 863
## hu_last_born1 0.04 0.06 -0.08 0.16 3000
## Rhat
## Intercept 1.00
## paternalage 1.01
## birth.cohort1760M1765 1.00
## birth.cohort1765M1770 1.00
## birth.cohort1770M1775 1.00
## birth.cohort1775M1780 1.00
## birth.cohort1780M1785 1.00
## birth.cohort1785M1790 1.00
## birth.cohort1790M1795 1.01
## birth.cohort1795M1800 1.00
## birth.cohort1800M1805 1.00
## birth.cohort1805M1810 1.00
## birth.cohort1810M1815 1.00
## birth.cohort1815M1820 1.01
## birth.cohort1820M1825 1.00
## birth.cohort1825M1830 1.00
## birth.cohort1830M1835 1.00
## first_born_adult_male 1.00
## last_born_adult_male 1.00
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.01
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.01
## older_siblings3 1.01
## older_siblings4 1.01
## older_siblings5P 1.01
## nr.siblings 1.00
## last_born1 1.00
## hu_Intercept 1.00
## hu_paternalage 1.00
## hu_birth.cohort1760M1765 1.00
## hu_birth.cohort1765M1770 1.00
## hu_birth.cohort1770M1775 1.00
## hu_birth.cohort1775M1780 1.00
## hu_birth.cohort1780M1785 1.00
## hu_birth.cohort1785M1790 1.00
## hu_birth.cohort1790M1795 1.00
## hu_birth.cohort1795M1800 1.00
## hu_birth.cohort1800M1805 1.00
## hu_birth.cohort1805M1810 1.00
## hu_birth.cohort1810M1815 1.01
## hu_birth.cohort1815M1820 1.00
## hu_birth.cohort1820M1825 1.01
## hu_birth.cohort1825M1830 1.00
## hu_birth.cohort1830M1835 1.00
## hu_first_born_adult_male 1.00
## hu_last_born_adult_male 1.00
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.00
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.01
## hu_older_siblings3 1.01
## hu_older_siblings4 1.00
## hu_older_siblings5P 1.01
## hu_nr.siblings 1.00
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.35 | 4.576 | 6.243 |
paternalage | 1.088 | 0.9814 | 1.211 |
birth.cohort1760M1765 | 0.9965 | 0.8802 | 1.135 |
birth.cohort1765M1770 | 0.8865 | 0.7928 | 0.9944 |
birth.cohort1770M1775 | 0.8879 | 0.7931 | 0.9976 |
birth.cohort1775M1780 | 0.9739 | 0.8766 | 1.081 |
birth.cohort1780M1785 | 0.8927 | 0.7959 | 0.9989 |
birth.cohort1785M1790 | 0.9084 | 0.815 | 1.013 |
birth.cohort1790M1795 | 0.9255 | 0.8336 | 1.027 |
birth.cohort1795M1800 | 0.8987 | 0.8178 | 0.9896 |
birth.cohort1800M1805 | 0.8927 | 0.811 | 0.98 |
birth.cohort1805M1810 | 0.8718 | 0.795 | 0.9557 |
birth.cohort1810M1815 | 0.9056 | 0.8262 | 0.9951 |
birth.cohort1815M1820 | 0.8671 | 0.7947 | 0.9485 |
birth.cohort1820M1825 | 0.8308 | 0.758 | 0.9099 |
birth.cohort1825M1830 | 0.8101 | 0.7432 | 0.8881 |
birth.cohort1830M1835 | 0.8312 | 0.7611 | 0.9095 |
first_born_adult_male | 0.9701 | 0.9231 | 1.021 |
last_born_adult_male | 0.9465 | 0.8987 | 0.9974 |
male1 | 1.125 | 1.07 | 1.183 |
maternalage.factor1420 | 0.9514 | 0.7924 | 1.13 |
maternalage.factor3550 | 0.9996 | 0.9481 | 1.057 |
paternalage.mean | 0.9161 | 0.8201 | 1.015 |
paternal_loss01 | 0.8642 | 0.7457 | 1.003 |
paternal_loss15 | 0.9667 | 0.8742 | 1.069 |
paternal_loss510 | 0.9338 | 0.8599 | 1.018 |
paternal_loss1015 | 1.008 | 0.9343 | 1.089 |
paternal_loss1520 | 0.9103 | 0.8418 | 0.9809 |
paternal_loss2025 | 0.8857 | 0.8219 | 0.9543 |
paternal_loss2530 | 0.9909 | 0.9262 | 1.061 |
paternal_loss3035 | 0.9715 | 0.9116 | 1.038 |
paternal_loss3540 | 0.9903 | 0.9299 | 1.054 |
paternal_loss4045 | 0.9885 | 0.9222 | 1.063 |
maternal_loss01 | 1.113 | 0.9518 | 1.3 |
maternal_loss15 | 0.9804 | 0.8908 | 1.077 |
maternal_loss510 | 1.074 | 0.9917 | 1.162 |
maternal_loss1015 | 1.028 | 0.9483 | 1.114 |
maternal_loss1520 | 1.001 | 0.9226 | 1.083 |
maternal_loss2025 | 1.002 | 0.9296 | 1.081 |
maternal_loss2530 | 0.9765 | 0.913 | 1.043 |
maternal_loss3035 | 0.9484 | 0.8882 | 1.011 |
maternal_loss3540 | 0.9639 | 0.9107 | 1.022 |
maternal_loss4045 | 0.9705 | 0.9108 | 1.031 |
older_siblings1 | 1.017 | 0.9632 | 1.072 |
older_siblings2 | 0.9415 | 0.8748 | 1.013 |
older_siblings3 | 0.9139 | 0.8289 | 1.004 |
older_siblings4 | 0.896 | 0.7972 | 1.005 |
older_siblings5P | 0.8892 | 0.7634 | 1.035 |
nr.siblings | 1.009 | 0.9945 | 1.024 |
last_born1 | 0.9551 | 0.9132 | 1 |
hu_Intercept | 1.041 | 0.6911 | 1.553 |
hu_paternalage | 1.478 | 1.106 | 1.972 |
hu_birth.cohort1760M1765 | 0.9807 | 0.7028 | 1.389 |
hu_birth.cohort1765M1770 | 0.7332 | 0.5492 | 0.9762 |
hu_birth.cohort1770M1775 | 0.9056 | 0.6822 | 1.226 |
hu_birth.cohort1775M1780 | 0.7796 | 0.5867 | 1.048 |
hu_birth.cohort1780M1785 | 0.7235 | 0.5341 | 0.9762 |
hu_birth.cohort1785M1790 | 0.6471 | 0.4882 | 0.8643 |
hu_birth.cohort1790M1795 | 0.7177 | 0.5442 | 0.9412 |
hu_birth.cohort1795M1800 | 0.6271 | 0.4904 | 0.8193 |
hu_birth.cohort1800M1805 | 0.5825 | 0.4549 | 0.7535 |
hu_birth.cohort1805M1810 | 0.7531 | 0.5873 | 0.9677 |
hu_birth.cohort1810M1815 | 0.6483 | 0.5077 | 0.8282 |
hu_birth.cohort1815M1820 | 0.4864 | 0.3859 | 0.6233 |
hu_birth.cohort1820M1825 | 0.5819 | 0.4574 | 0.7354 |
hu_birth.cohort1825M1830 | 0.5895 | 0.4642 | 0.7488 |
hu_birth.cohort1830M1835 | 0.5681 | 0.454 | 0.7328 |
hu_first_born_adult_male | 0.3541 | 0.3075 | 0.4074 |
hu_last_born_adult_male | 0.3901 | 0.3353 | 0.4499 |
hu_male1 | 2.736 | 2.425 | 3.087 |
hu_maternalage.factor1420 | 1.287 | 0.8142 | 2.049 |
hu_maternalage.factor3550 | 1.18 | 1.013 | 1.365 |
hu_paternalage.mean | 0.7333 | 0.5475 | 0.981 |
hu_paternal_loss01 | 1.836 | 1.269 | 2.67 |
hu_paternal_loss15 | 1.764 | 1.37 | 2.284 |
hu_paternal_loss510 | 1.212 | 0.9749 | 1.513 |
hu_paternal_loss1015 | 1.163 | 0.9402 | 1.436 |
hu_paternal_loss1520 | 1.084 | 0.8858 | 1.323 |
hu_paternal_loss2025 | 1.126 | 0.9253 | 1.36 |
hu_paternal_loss2530 | 1.064 | 0.8804 | 1.278 |
hu_paternal_loss3035 | 0.966 | 0.8059 | 1.157 |
hu_paternal_loss3540 | 0.9843 | 0.8264 | 1.174 |
hu_paternal_loss4045 | 1.12 | 0.9149 | 1.375 |
hu_maternal_loss01 | 4.867 | 3.374 | 7.12 |
hu_maternal_loss15 | 1.814 | 1.42 | 2.285 |
hu_maternal_loss510 | 1.639 | 1.324 | 2.017 |
hu_maternal_loss1015 | 1.628 | 1.311 | 2.032 |
hu_maternal_loss1520 | 1.351 | 1.092 | 1.685 |
hu_maternal_loss2025 | 1.274 | 1.031 | 1.581 |
hu_maternal_loss2530 | 1.197 | 0.9895 | 1.45 |
hu_maternal_loss3035 | 1.237 | 1.028 | 1.48 |
hu_maternal_loss3540 | 1.031 | 0.877 | 1.206 |
hu_maternal_loss4045 | 1.342 | 1.131 | 1.611 |
hu_older_siblings1 | 0.8529 | 0.7302 | 0.9923 |
hu_older_siblings2 | 0.6886 | 0.5638 | 0.8404 |
hu_older_siblings3 | 0.6322 | 0.4889 | 0.8206 |
hu_older_siblings4 | 0.6145 | 0.4437 | 0.8401 |
hu_older_siblings5P | 0.4278 | 0.2864 | 0.6503 |
hu_nr.siblings | 1.074 | 1.033 | 1.117 |
hu_last_born1 | 1.042 | 0.9231 | 1.177 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.64 | [2.34;2.95] | [2.44;2.85] |
estimate father 35y | 2.43 | [2.04;2.9] | [2.17;2.72] |
percentage change | -7.93 | [-21.26;8.17] | [-17.12;2.35] |
OR/IRR | 1.09 | [0.98;1.21] | [1.02;1.17] |
OR hurdle | 1.47 | [1.11;1.97] | [1.22;1.8] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r8_adjust_for_first_born_adult.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
In our main model, we control for birth cohort in 5-year-bins (lumping small bins). We chose to do so, because nonlinear and even sharply spiking effects of birth cohort are plausible (due to e.g. epidemics). This decision may be disputed, as it summarises 5-year-bins. Here, we instead allow for a thin-splate spline on the continuous birth year variable. This allows for smooth nonlinear (but not spiking) birth cohort effects.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
## Warning: There were 9 divergent transitions after warmup. Increasing
## adapt_delta above 0.8 may help. See http://mc-stan.org/misc/
## warnings.html#divergent-transitions-after-warmup
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + s(byear) + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + s(byear) + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## sds ~ student_t(3, 0, 10)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
## sds_hu ~ student_t(3, 0, 10)
##
## Smooth Terms:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sds(sbyear_1) 0.12 0.13 0.00 0.47 936 1.01
## sds(hu_sbyear_1) 0.31 0.35 0.01 1.27 829 1.01
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.22 0.01 0.20 0.25 998 1
## sd(hu_Intercept) 0.46 0.04 0.38 0.55 802 1
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.53 0.07 1.40 1.67 2666
## paternalage 0.07 0.06 -0.04 0.17 761
## male1 0.08 0.02 0.05 0.11 3000
## maternalage.factor1420 -0.05 0.09 -0.24 0.13 3000
## maternalage.factor3550 0.00 0.03 -0.06 0.05 3000
## paternalage.mean -0.07 0.06 -0.18 0.04 827
## paternal_loss01 -0.15 0.07 -0.30 -0.01 3000
## paternal_loss15 -0.04 0.05 -0.14 0.06 3000
## paternal_loss510 -0.07 0.04 -0.15 0.01 1607
## paternal_loss1015 0.00 0.04 -0.07 0.08 1534
## paternal_loss1520 -0.09 0.04 -0.17 -0.02 1597
## paternal_loss2025 -0.12 0.04 -0.20 -0.05 1554
## paternal_loss2530 -0.01 0.03 -0.08 0.05 1248
## paternal_loss3035 -0.03 0.03 -0.09 0.04 1440
## paternal_loss3540 -0.01 0.03 -0.07 0.05 1508
## paternal_loss4045 -0.01 0.04 -0.09 0.06 1743
## maternal_loss01 0.11 0.08 -0.05 0.26 3000
## maternal_loss15 -0.02 0.05 -0.10 0.07 3000
## maternal_loss510 0.07 0.04 -0.01 0.15 2031
## maternal_loss1015 0.03 0.04 -0.05 0.11 2282
## maternal_loss1520 0.00 0.04 -0.08 0.08 3000
## maternal_loss2025 0.00 0.04 -0.07 0.08 3000
## maternal_loss2530 -0.02 0.03 -0.09 0.04 3000
## maternal_loss3035 -0.05 0.03 -0.12 0.01 3000
## maternal_loss3540 -0.04 0.03 -0.09 0.02 3000
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## older_siblings1 0.03 0.03 -0.03 0.08 1475
## older_siblings2 -0.05 0.04 -0.12 0.02 874
## older_siblings3 -0.07 0.05 -0.17 0.02 845
## older_siblings4 -0.10 0.06 -0.21 0.02 875
## older_siblings5P -0.10 0.08 -0.25 0.06 753
## nr.siblings 0.01 0.01 0.00 0.02 1084
## last_born1 -0.04 0.02 -0.09 0.00 3000
## sbyear_1 -0.05 0.05 -0.15 0.04 895
## hu_Intercept -0.72 0.17 -1.04 -0.38 3000
## hu_paternalage 0.27 0.14 -0.01 0.56 749
## hu_male1 0.27 0.05 0.18 0.37 3000
## hu_maternalage.factor1420 0.25 0.23 -0.20 0.69 3000
## hu_maternalage.factor3550 0.13 0.07 -0.01 0.27 3000
## hu_paternalage.mean -0.19 0.15 -0.48 0.10 768
## hu_paternal_loss01 0.57 0.18 0.24 0.92 3000
## hu_paternal_loss15 0.53 0.13 0.28 0.78 3000
## hu_paternal_loss510 0.19 0.11 -0.03 0.42 1553
## hu_paternal_loss1015 0.14 0.11 -0.07 0.35 1864
## hu_paternal_loss1520 0.10 0.10 -0.10 0.30 1318
## hu_paternal_loss2025 0.16 0.10 -0.03 0.35 1744
## hu_paternal_loss2530 0.06 0.09 -0.12 0.23 1328
## hu_paternal_loss3035 -0.02 0.09 -0.20 0.15 1536
## hu_paternal_loss3540 -0.02 0.09 -0.19 0.16 1530
## hu_paternal_loss4045 0.13 0.10 -0.06 0.32 3000
## hu_maternal_loss01 1.59 0.19 1.22 1.97 3000
## hu_maternal_loss15 0.60 0.12 0.36 0.84 3000
## hu_maternal_loss510 0.48 0.11 0.27 0.69 3000
## hu_maternal_loss1015 0.46 0.11 0.24 0.67 3000
## hu_maternal_loss1520 0.31 0.11 0.09 0.53 3000
## hu_maternal_loss2025 0.27 0.10 0.08 0.46 3000
## hu_maternal_loss2530 0.19 0.09 0.01 0.37 3000
## hu_maternal_loss3035 0.22 0.09 0.04 0.38 3000
## hu_maternal_loss3540 0.06 0.08 -0.09 0.22 3000
## hu_maternal_loss4045 0.26 0.09 0.10 0.43 3000
## hu_older_siblings1 -0.01 0.08 -0.16 0.14 1389
## hu_older_siblings2 -0.15 0.10 -0.35 0.03 912
## hu_older_siblings3 -0.19 0.13 -0.44 0.06 787
## hu_older_siblings4 -0.23 0.16 -0.53 0.07 747
## hu_older_siblings5P -0.57 0.21 -0.97 -0.16 717
## hu_nr.siblings 0.10 0.02 0.07 0.14 888
## hu_last_born1 0.07 0.06 -0.04 0.20 3000
## hu_sbyear_1 -0.15 0.11 -0.37 0.09 1269
## Rhat
## Intercept 1.00
## paternalage 1.01
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.01
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.01
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.00
## older_siblings3 1.00
## older_siblings4 1.00
## older_siblings5P 1.01
## nr.siblings 1.01
## last_born1 1.00
## sbyear_1 1.01
## hu_Intercept 1.00
## hu_paternalage 1.00
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.00
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.01
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.00
## hu_older_siblings3 1.00
## hu_older_siblings4 1.00
## hu_older_siblings5P 1.00
## hu_nr.siblings 1.00
## hu_last_born1 1.00
## hu_sbyear_1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 4.635 | 4.06 | 5.3 |
paternalage | 1.069 | 0.9609 | 1.19 |
male1 | 1.081 | 1.049 | 1.116 |
maternalage.factor1420 | 0.9467 | 0.7855 | 1.135 |
maternalage.factor3550 | 0.9986 | 0.9441 | 1.056 |
paternalage.mean | 0.9345 | 0.8375 | 1.046 |
paternal_loss01 | 0.8579 | 0.7392 | 0.9912 |
paternal_loss15 | 0.9634 | 0.8734 | 1.066 |
paternal_loss510 | 0.9321 | 0.8598 | 1.008 |
paternal_loss1015 | 1.002 | 0.928 | 1.083 |
paternal_loss1520 | 0.9102 | 0.8469 | 0.9845 |
paternal_loss2025 | 0.8841 | 0.8206 | 0.9527 |
paternal_loss2530 | 0.9863 | 0.9232 | 1.053 |
paternal_loss3035 | 0.9706 | 0.9115 | 1.036 |
paternal_loss3540 | 0.988 | 0.9287 | 1.053 |
paternal_loss4045 | 0.9881 | 0.9181 | 1.061 |
maternal_loss01 | 1.111 | 0.9498 | 1.297 |
maternal_loss15 | 0.9835 | 0.9014 | 1.072 |
maternal_loss510 | 1.069 | 0.986 | 1.157 |
maternal_loss1015 | 1.031 | 0.9495 | 1.118 |
maternal_loss1520 | 1.001 | 0.9235 | 1.082 |
maternal_loss2025 | 1.004 | 0.9332 | 1.079 |
maternal_loss2530 | 0.9772 | 0.9118 | 1.045 |
maternal_loss3035 | 0.9486 | 0.8885 | 1.01 |
maternal_loss3540 | 0.9634 | 0.9097 | 1.02 |
maternal_loss4045 | 0.9676 | 0.9098 | 1.029 |
older_siblings1 | 1.026 | 0.9724 | 1.081 |
older_siblings2 | 0.9521 | 0.8854 | 1.022 |
older_siblings3 | 0.9304 | 0.8478 | 1.022 |
older_siblings4 | 0.909 | 0.8081 | 1.025 |
older_siblings5P | 0.9066 | 0.7752 | 1.066 |
nr.siblings | 1.01 | 0.9951 | 1.024 |
last_born1 | 0.9585 | 0.9155 | 1.004 |
sbyear_1 | 0.9467 | 0.8616 | 1.039 |
hu_Intercept | 0.488 | 0.3517 | 0.683 |
hu_paternalage | 1.307 | 0.9897 | 1.745 |
hu_male1 | 1.316 | 1.203 | 1.442 |
hu_maternalage.factor1420 | 1.279 | 0.8223 | 1.995 |
hu_maternalage.factor3550 | 1.141 | 0.9919 | 1.315 |
hu_paternalage.mean | 0.8295 | 0.6213 | 1.109 |
hu_paternal_loss01 | 1.772 | 1.272 | 2.512 |
hu_paternal_loss15 | 1.698 | 1.323 | 2.173 |
hu_paternal_loss510 | 1.21 | 0.9734 | 1.515 |
hu_paternal_loss1015 | 1.152 | 0.9303 | 1.419 |
hu_paternal_loss1520 | 1.107 | 0.9046 | 1.35 |
hu_paternal_loss2025 | 1.168 | 0.9697 | 1.418 |
hu_paternal_loss2530 | 1.059 | 0.8864 | 1.264 |
hu_paternal_loss3035 | 0.9777 | 0.8198 | 1.167 |
hu_paternal_loss3540 | 0.9812 | 0.8263 | 1.177 |
hu_paternal_loss4045 | 1.135 | 0.9393 | 1.374 |
hu_maternal_loss01 | 4.885 | 3.393 | 7.165 |
hu_maternal_loss15 | 1.822 | 1.438 | 2.306 |
hu_maternal_loss510 | 1.615 | 1.315 | 1.989 |
hu_maternal_loss1015 | 1.579 | 1.271 | 1.945 |
hu_maternal_loss1520 | 1.357 | 1.094 | 1.691 |
hu_maternal_loss2025 | 1.311 | 1.084 | 1.591 |
hu_maternal_loss2530 | 1.21 | 1.013 | 1.446 |
hu_maternal_loss3035 | 1.243 | 1.044 | 1.467 |
hu_maternal_loss3540 | 1.067 | 0.9116 | 1.241 |
hu_maternal_loss4045 | 1.296 | 1.102 | 1.53 |
hu_older_siblings1 | 0.9886 | 0.8512 | 1.147 |
hu_older_siblings2 | 0.8586 | 0.7031 | 1.032 |
hu_older_siblings3 | 0.8251 | 0.6416 | 1.063 |
hu_older_siblings4 | 0.7922 | 0.5857 | 1.076 |
hu_older_siblings5P | 0.5644 | 0.3792 | 0.8493 |
hu_nr.siblings | 1.11 | 1.071 | 1.152 |
hu_last_born1 | 1.078 | 0.9591 | 1.224 |
hu_sbyear_1 | 0.8597 | 0.693 | 1.095 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.54 | [2.28;2.81] | [2.37;2.72] |
estimate father 35y | 2.38 | [2.01;2.78] | [2.13;2.63] |
percentage change | -6.35 | [-21.47;11.08] | [-16.4;4.87] |
OR/IRR | 1.07 | [0.96;1.19] | [1;1.15] |
OR hurdle | 1.31 | [0.99;1.74] | [1.08;1.58] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r9_continuous_byear_adjustment.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
Paternal age effects may vary between different families. Although we did not explore between-family moderators of paternal age effects in our study, we tested whether modelling an additional group-level slope for paternal age differences within the family, would change the results by allowing for shrinkage and to examine the amount of inter-family differences to be explained for potential future moderator analysis.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
## Warning: There were 3 divergent transitions after warmup. Increasing
## adapt_delta above 0.8 may help. See http://mc-stan.org/misc/
## warnings.html#divergent-transitions-after-warmup
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 + paternalage | idParents)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 + paternalage | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 10000; warmup = 4000; thin = 5;
## total post-warmup samples = 7200
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## L ~ lkj_corr_cholesky(1)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI
## sd(Intercept) 0.71 0.09 0.52 0.88
## sd(paternalage) 0.18 0.03 0.12 0.23
## sd(hu_Intercept) 0.38 0.19 0.03 0.78
## sd(hu_paternalage) 0.09 0.05 0.01 0.21
## cor(Intercept,paternalage) -0.96 0.02 -0.98 -0.92
## cor(hu_Intercept,hu_paternalage) -0.07 0.54 -0.89 0.92
## Eff.Sample Rhat
## sd(Intercept) 3055 1.00
## sd(paternalage) 2382 1.00
## sd(hu_Intercept) 1623 1.00
## sd(hu_paternalage) 375 1.01
## cor(Intercept,paternalage) 3469 1.00
## cor(hu_Intercept,hu_paternalage) 989 1.00
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.63 0.08 1.47 1.80 6098
## paternalage 0.07 0.06 -0.04 0.19 5432
## birth_cohort1760M1765 -0.01 0.06 -0.14 0.12 6137
## birth_cohort1765M1770 -0.12 0.06 -0.23 -0.01 5509
## birth_cohort1770M1775 -0.12 0.06 -0.23 0.00 5569
## birth_cohort1775M1780 -0.04 0.06 -0.15 0.07 4964
## birth_cohort1780M1785 -0.12 0.06 -0.23 0.00 5773
## birth_cohort1785M1790 -0.11 0.06 -0.22 0.00 4841
## birth_cohort1790M1795 -0.08 0.05 -0.18 0.03 4610
## birth_cohort1795M1800 -0.11 0.05 -0.21 -0.01 5465
## birth_cohort1800M1805 -0.12 0.05 -0.22 -0.02 5205
## birth_cohort1805M1810 -0.14 0.05 -0.24 -0.04 5154
## birth_cohort1810M1815 -0.10 0.05 -0.20 -0.01 4755
## birth_cohort1815M1820 -0.15 0.05 -0.24 -0.06 4574
## birth_cohort1820M1825 -0.20 0.05 -0.29 -0.11 4256
## birth_cohort1825M1830 -0.22 0.05 -0.31 -0.13 4797
## birth_cohort1830M1835 -0.19 0.05 -0.28 -0.09 4854
## male1 0.08 0.02 0.05 0.11 7028
## maternalage.factor1420 -0.05 0.10 -0.25 0.14 7200
## maternalage.factor3550 -0.01 0.03 -0.06 0.05 7200
## paternalage.mean -0.07 0.06 -0.19 0.04 5637
## paternal_loss01 -0.15 0.08 -0.30 0.00 7200
## paternal_loss15 -0.04 0.05 -0.14 0.06 6636
## paternal_loss510 -0.08 0.04 -0.16 0.01 6086
## paternal_loss1015 0.00 0.04 -0.08 0.08 6556
## paternal_loss1520 -0.09 0.04 -0.17 -0.01 6344
## paternal_loss2025 -0.12 0.04 -0.19 -0.04 5556
## paternal_loss2530 0.00 0.04 -0.08 0.06 6421
## paternal_loss3035 -0.02 0.03 -0.09 0.04 6486
## paternal_loss3540 -0.01 0.03 -0.07 0.06 6498
## paternal_loss4045 -0.01 0.04 -0.08 0.06 6688
## maternal_loss01 0.10 0.08 -0.06 0.25 6426
## maternal_loss15 -0.01 0.05 -0.11 0.08 6522
## maternal_loss510 0.08 0.04 0.00 0.16 6995
## maternal_loss1015 0.04 0.04 -0.05 0.12 7200
## maternal_loss1520 0.00 0.04 -0.08 0.09 6247
## maternal_loss2025 0.01 0.04 -0.06 0.09 7200
## maternal_loss2530 -0.02 0.04 -0.09 0.05 6373
## maternal_loss3035 -0.04 0.03 -0.11 0.02 7042
## maternal_loss3540 -0.03 0.03 -0.08 0.03 6626
## maternal_loss4045 -0.03 0.03 -0.09 0.04 6922
## older_siblings1 0.03 0.03 -0.03 0.08 6499
## older_siblings2 -0.04 0.04 -0.12 0.03 5862
## older_siblings3 -0.06 0.05 -0.16 0.03 5457
## older_siblings4 -0.09 0.06 -0.21 0.03 5365
## older_siblings5P -0.09 0.08 -0.26 0.07 5108
## nr.siblings 0.01 0.01 0.00 0.02 5755
## last_born1 -0.04 0.02 -0.09 0.00 7200
## hu_Intercept -0.33 0.20 -0.73 0.04 5660
## hu_paternalage 0.27 0.14 0.00 0.56 5204
## hu_birth_cohort1760M1765 -0.05 0.17 -0.37 0.29 6156
## hu_birth_cohort1765M1770 -0.31 0.15 -0.60 -0.01 5878
## hu_birth_cohort1770M1775 -0.07 0.15 -0.35 0.22 5488
## hu_birth_cohort1775M1780 -0.20 0.14 -0.47 0.09 5508
## hu_birth_cohort1780M1785 -0.27 0.15 -0.56 0.01 5653
## hu_birth_cohort1785M1790 -0.42 0.14 -0.70 -0.13 5560
## hu_birth_cohort1790M1795 -0.31 0.13 -0.57 -0.05 5122
## hu_birth_cohort1795M1800 -0.45 0.13 -0.70 -0.19 5377
## hu_birth_cohort1800M1805 -0.53 0.12 -0.78 -0.29 5514
## hu_birth_cohort1805M1810 -0.26 0.13 -0.51 -0.02 5520
## hu_birth_cohort1810M1815 -0.44 0.12 -0.67 -0.20 5187
## hu_birth_cohort1815M1820 -0.70 0.12 -0.93 -0.47 5205
## hu_birth_cohort1820M1825 -0.52 0.12 -0.75 -0.29 5218
## hu_birth_cohort1825M1830 -0.54 0.12 -0.78 -0.31 4984
## hu_birth_cohort1830M1835 -0.56 0.12 -0.79 -0.32 5239
## hu_male1 0.27 0.05 0.19 0.36 6970
## hu_maternalage.factor1420 0.25 0.23 -0.19 0.72 7200
## hu_maternalage.factor3550 0.13 0.07 -0.01 0.27 7200
## hu_paternalage.mean -0.19 0.15 -0.48 0.10 5314
## hu_paternal_loss01 0.57 0.18 0.21 0.93 6845
## hu_paternal_loss15 0.54 0.13 0.29 0.79 6671
## hu_paternal_loss510 0.19 0.11 -0.03 0.41 6421
## hu_paternal_loss1015 0.16 0.10 -0.05 0.36 6184
## hu_paternal_loss1520 0.10 0.10 -0.10 0.30 6605
## hu_paternal_loss2025 0.15 0.10 -0.03 0.35 6452
## hu_paternal_loss2530 0.06 0.09 -0.12 0.24 6379
## hu_paternal_loss3035 -0.02 0.09 -0.20 0.15 6013
## hu_paternal_loss3540 -0.02 0.09 -0.19 0.16 6448
## hu_paternal_loss4045 0.14 0.10 -0.05 0.34 7012
## hu_maternal_loss01 1.58 0.19 1.21 1.95 6948
## hu_maternal_loss15 0.58 0.12 0.35 0.82 6888
## hu_maternal_loss510 0.48 0.11 0.27 0.69 6840
## hu_maternal_loss1015 0.46 0.11 0.25 0.68 7008
## hu_maternal_loss1520 0.30 0.11 0.09 0.52 7055
## hu_maternal_loss2025 0.26 0.10 0.06 0.47 7119
## hu_maternal_loss2530 0.19 0.09 0.01 0.37 7200
## hu_maternal_loss3035 0.22 0.09 0.05 0.39 6972
## hu_maternal_loss3540 0.07 0.08 -0.10 0.22 7200
## hu_maternal_loss4045 0.28 0.08 0.11 0.44 6930
## hu_older_siblings1 -0.01 0.07 -0.16 0.13 6576
## hu_older_siblings2 -0.16 0.10 -0.35 0.03 5814
## hu_older_siblings3 -0.20 0.13 -0.45 0.04 5382
## hu_older_siblings4 -0.24 0.16 -0.54 0.07 5809
## hu_older_siblings5P -0.58 0.20 -0.98 -0.18 5468
## hu_nr.siblings 0.10 0.02 0.07 0.14 5772
## hu_last_born1 0.08 0.06 -0.04 0.21 6346
## Rhat
## Intercept 1
## paternalage 1
## birth_cohort1760M1765 1
## birth_cohort1765M1770 1
## birth_cohort1770M1775 1
## birth_cohort1775M1780 1
## birth_cohort1780M1785 1
## birth_cohort1785M1790 1
## birth_cohort1790M1795 1
## birth_cohort1795M1800 1
## birth_cohort1800M1805 1
## birth_cohort1805M1810 1
## birth_cohort1810M1815 1
## birth_cohort1815M1820 1
## birth_cohort1820M1825 1
## birth_cohort1825M1830 1
## birth_cohort1830M1835 1
## male1 1
## maternalage.factor1420 1
## maternalage.factor3550 1
## paternalage.mean 1
## paternal_loss01 1
## paternal_loss15 1
## paternal_loss510 1
## paternal_loss1015 1
## paternal_loss1520 1
## paternal_loss2025 1
## paternal_loss2530 1
## paternal_loss3035 1
## paternal_loss3540 1
## paternal_loss4045 1
## maternal_loss01 1
## maternal_loss15 1
## maternal_loss510 1
## maternal_loss1015 1
## maternal_loss1520 1
## maternal_loss2025 1
## maternal_loss2530 1
## maternal_loss3035 1
## maternal_loss3540 1
## maternal_loss4045 1
## older_siblings1 1
## older_siblings2 1
## older_siblings3 1
## older_siblings4 1
## older_siblings5P 1
## nr.siblings 1
## last_born1 1
## hu_Intercept 1
## hu_paternalage 1
## hu_birth_cohort1760M1765 1
## hu_birth_cohort1765M1770 1
## hu_birth_cohort1770M1775 1
## hu_birth_cohort1775M1780 1
## hu_birth_cohort1780M1785 1
## hu_birth_cohort1785M1790 1
## hu_birth_cohort1790M1795 1
## hu_birth_cohort1795M1800 1
## hu_birth_cohort1800M1805 1
## hu_birth_cohort1805M1810 1
## hu_birth_cohort1810M1815 1
## hu_birth_cohort1815M1820 1
## hu_birth_cohort1820M1825 1
## hu_birth_cohort1825M1830 1
## hu_birth_cohort1830M1835 1
## hu_male1 1
## hu_maternalage.factor1420 1
## hu_maternalage.factor3550 1
## hu_paternalage.mean 1
## hu_paternal_loss01 1
## hu_paternal_loss15 1
## hu_paternal_loss510 1
## hu_paternal_loss1015 1
## hu_paternal_loss1520 1
## hu_paternal_loss2025 1
## hu_paternal_loss2530 1
## hu_paternal_loss3035 1
## hu_paternal_loss3540 1
## hu_paternal_loss4045 1
## hu_maternal_loss01 1
## hu_maternal_loss15 1
## hu_maternal_loss510 1
## hu_maternal_loss1015 1
## hu_maternal_loss1520 1
## hu_maternal_loss2025 1
## hu_maternal_loss2530 1
## hu_maternal_loss3035 1
## hu_maternal_loss3540 1
## hu_maternal_loss4045 1
## hu_older_siblings1 1
## hu_older_siblings2 1
## hu_older_siblings3 1
## hu_older_siblings4 1
## hu_older_siblings5P 1
## hu_nr.siblings 1
## hu_last_born1 1
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.117 | 4.351 | 6.029 |
paternalage | 1.076 | 0.9647 | 1.204 |
birth_cohort1760M1765 | 0.991 | 0.8713 | 1.123 |
birth_cohort1765M1770 | 0.8854 | 0.7909 | 0.993 |
birth_cohort1770M1775 | 0.8904 | 0.7913 | 1.001 |
birth_cohort1775M1780 | 0.9581 | 0.8586 | 1.071 |
birth_cohort1780M1785 | 0.8898 | 0.7926 | 0.9958 |
birth_cohort1785M1790 | 0.8946 | 0.801 | 0.9985 |
birth_cohort1790M1795 | 0.9252 | 0.8324 | 1.027 |
birth_cohort1795M1800 | 0.8941 | 0.809 | 0.9862 |
birth_cohort1800M1805 | 0.8893 | 0.8053 | 0.9775 |
birth_cohort1805M1810 | 0.8709 | 0.7872 | 0.9613 |
birth_cohort1810M1815 | 0.9013 | 0.8187 | 0.9918 |
birth_cohort1815M1820 | 0.86 | 0.7852 | 0.9415 |
birth_cohort1820M1825 | 0.8206 | 0.747 | 0.8987 |
birth_cohort1825M1830 | 0.8032 | 0.7318 | 0.88 |
birth_cohort1830M1835 | 0.8299 | 0.7545 | 0.9109 |
male1 | 1.081 | 1.047 | 1.118 |
maternalage.factor1420 | 0.9483 | 0.7823 | 1.146 |
maternalage.factor3550 | 0.9919 | 0.9387 | 1.048 |
paternalage.mean | 0.9318 | 0.8299 | 1.04 |
paternal_loss01 | 0.8615 | 0.7429 | 0.9958 |
paternal_loss15 | 0.9613 | 0.8698 | 1.062 |
paternal_loss510 | 0.925 | 0.8489 | 1.01 |
paternal_loss1015 | 1.002 | 0.9258 | 1.083 |
paternal_loss1520 | 0.9103 | 0.8423 | 0.9857 |
paternal_loss2025 | 0.8868 | 0.8233 | 0.9567 |
paternal_loss2530 | 0.9951 | 0.9274 | 1.067 |
paternal_loss3035 | 0.9756 | 0.9128 | 1.044 |
paternal_loss3540 | 0.9912 | 0.9287 | 1.057 |
paternal_loss4045 | 0.9912 | 0.9219 | 1.066 |
maternal_loss01 | 1.103 | 0.9401 | 1.289 |
maternal_loss15 | 0.9862 | 0.8979 | 1.087 |
maternal_loss510 | 1.08 | 0.9965 | 1.171 |
maternal_loss1015 | 1.037 | 0.9545 | 1.123 |
maternal_loss1520 | 1.004 | 0.9234 | 1.09 |
maternal_loss2025 | 1.013 | 0.9374 | 1.095 |
maternal_loss2530 | 0.9832 | 0.917 | 1.056 |
maternal_loss3035 | 0.9587 | 0.8985 | 1.023 |
maternal_loss3540 | 0.9746 | 0.9194 | 1.035 |
maternal_loss4045 | 0.9746 | 0.9145 | 1.038 |
older_siblings1 | 1.029 | 0.9738 | 1.086 |
older_siblings2 | 0.9579 | 0.8896 | 1.031 |
older_siblings3 | 0.9373 | 0.8518 | 1.03 |
older_siblings4 | 0.9177 | 0.8132 | 1.033 |
older_siblings5P | 0.9136 | 0.7747 | 1.068 |
nr.siblings | 1.01 | 0.9955 | 1.025 |
last_born1 | 0.9572 | 0.9129 | 1.004 |
hu_Intercept | 0.7155 | 0.4833 | 1.045 |
hu_paternalage | 1.316 | 0.9956 | 1.751 |
hu_birth_cohort1760M1765 | 0.9507 | 0.6876 | 1.335 |
hu_birth_cohort1765M1770 | 0.7343 | 0.5506 | 0.9873 |
hu_birth_cohort1770M1775 | 0.9362 | 0.702 | 1.248 |
hu_birth_cohort1775M1780 | 0.8217 | 0.6232 | 1.093 |
hu_birth_cohort1780M1785 | 0.7604 | 0.5689 | 1.011 |
hu_birth_cohort1785M1790 | 0.6564 | 0.496 | 0.8757 |
hu_birth_cohort1790M1795 | 0.7349 | 0.5645 | 0.9556 |
hu_birth_cohort1795M1800 | 0.6352 | 0.4952 | 0.8237 |
hu_birth_cohort1800M1805 | 0.5868 | 0.4605 | 0.7502 |
hu_birth_cohort1805M1810 | 0.7673 | 0.5991 | 0.9803 |
hu_birth_cohort1810M1815 | 0.647 | 0.5104 | 0.8224 |
hu_birth_cohort1815M1820 | 0.4966 | 0.3926 | 0.626 |
hu_birth_cohort1820M1825 | 0.5953 | 0.472 | 0.7476 |
hu_birth_cohort1825M1830 | 0.5808 | 0.46 | 0.7335 |
hu_birth_cohort1830M1835 | 0.5736 | 0.4548 | 0.724 |
hu_male1 | 1.315 | 1.205 | 1.435 |
hu_maternalage.factor1420 | 1.286 | 0.8262 | 2.046 |
hu_maternalage.factor3550 | 1.14 | 0.9881 | 1.314 |
hu_paternalage.mean | 0.8258 | 0.6163 | 1.1 |
hu_paternal_loss01 | 1.775 | 1.239 | 2.538 |
hu_paternal_loss15 | 1.711 | 1.331 | 2.208 |
hu_paternal_loss510 | 1.213 | 0.9738 | 1.514 |
hu_paternal_loss1015 | 1.17 | 0.9544 | 1.438 |
hu_paternal_loss1520 | 1.109 | 0.9082 | 1.352 |
hu_paternal_loss2025 | 1.166 | 0.9656 | 1.412 |
hu_paternal_loss2530 | 1.063 | 0.8849 | 1.271 |
hu_paternal_loss3035 | 0.9787 | 0.8201 | 1.16 |
hu_paternal_loss3540 | 0.984 | 0.8261 | 1.171 |
hu_paternal_loss4045 | 1.152 | 0.9475 | 1.401 |
hu_maternal_loss01 | 4.845 | 3.369 | 7.003 |
hu_maternal_loss15 | 1.792 | 1.417 | 2.267 |
hu_maternal_loss510 | 1.618 | 1.312 | 1.992 |
hu_maternal_loss1015 | 1.591 | 1.286 | 1.967 |
hu_maternal_loss1520 | 1.356 | 1.097 | 1.677 |
hu_maternal_loss2025 | 1.295 | 1.06 | 1.592 |
hu_maternal_loss2530 | 1.208 | 1.01 | 1.446 |
hu_maternal_loss3035 | 1.241 | 1.05 | 1.472 |
hu_maternal_loss3540 | 1.068 | 0.9086 | 1.25 |
hu_maternal_loss4045 | 1.32 | 1.115 | 1.552 |
hu_older_siblings1 | 0.9873 | 0.8548 | 1.142 |
hu_older_siblings2 | 0.8541 | 0.7069 | 1.03 |
hu_older_siblings3 | 0.8196 | 0.6403 | 1.043 |
hu_older_siblings4 | 0.7899 | 0.5819 | 1.071 |
hu_older_siblings5P | 0.5613 | 0.3753 | 0.835 |
hu_nr.siblings | 1.111 | 1.071 | 1.153 |
hu_last_born1 | 1.085 | 0.9585 | 1.23 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.26 | [1.92;2.67] | [2.04;2.52] |
estimate father 35y | 2.07 | [1.64;2.57] | [1.78;2.39] |
percentage change | -8.56 | [-25.31;11.57] | [-19.96;3.93] |
OR/IRR | 1.08 | [0.96;1.2] | [1;1.16] |
OR hurdle | 1.32 | [1;1.75] | [1.09;1.58] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r10_add_random_slope.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
Most anchors in our sample are full biological siblings and especially in the historical populations, divorce and remarriage was rare. Therefore, we chose to include only one group-level effect, for the parent couple (i.e. one group-level effect per father-mother-dyad). Including one intercept per parent is potentially a better way to adjust for genetic propensities inherited from either parent and allows estimating this propensity also from half-siblings, while half-sibling relationships were ignored in our main models. This comes at the cost of modelling complexity.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idMere) + (1 | idPere)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idMere) + (1 | idPere)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 10000; warmup = 4000; thin = 5;
## total post-warmup samples = 7200
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idMere (Number of levels: 2124)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.16 0.05 0.03 0.23 361 1.01
## sd(hu_Intercept) 0.41 0.10 0.10 0.54 490 1.01
##
## ~idPere (Number of levels: 2038)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.13 0.06 0.01 0.23 345 1.01
## sd(hu_Intercept) 0.20 0.13 0.01 0.46 540 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.66 0.08 1.50 1.81 6344
## paternalage 0.07 0.05 -0.04 0.17 5378
## birth_cohort1760M1765 -0.01 0.06 -0.13 0.12 6228
## birth_cohort1765M1770 -0.12 0.06 -0.23 -0.01 5330
## birth_cohort1770M1775 -0.11 0.06 -0.23 0.00 5477
## birth_cohort1775M1780 -0.03 0.06 -0.13 0.08 5347
## birth_cohort1780M1785 -0.11 0.06 -0.22 0.00 5374
## birth_cohort1785M1790 -0.09 0.05 -0.20 0.01 5191
## birth_cohort1790M1795 -0.07 0.05 -0.18 0.03 5273
## birth_cohort1795M1800 -0.10 0.05 -0.20 -0.01 5346
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.02 4838
## birth_cohort1805M1810 -0.14 0.05 -0.23 -0.04 5037
## birth_cohort1810M1815 -0.10 0.05 -0.19 -0.01 4791
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 4789
## birth_cohort1820M1825 -0.18 0.05 -0.27 -0.10 5002
## birth_cohort1825M1830 -0.21 0.05 -0.30 -0.12 4956
## birth_cohort1830M1835 -0.18 0.05 -0.27 -0.09 5121
## male1 0.08 0.02 0.05 0.11 7049
## maternalage.factor1420 -0.06 0.09 -0.24 0.12 6924
## maternalage.factor3550 0.00 0.03 -0.06 0.05 6274
## paternalage.mean -0.07 0.06 -0.18 0.04 5465
## paternal_loss01 -0.15 0.07 -0.30 0.00 7014
## paternal_loss15 -0.03 0.05 -0.13 0.07 6696
## paternal_loss510 -0.07 0.04 -0.15 0.02 6569
## paternal_loss1015 0.01 0.04 -0.07 0.09 6457
## paternal_loss1520 -0.09 0.04 -0.17 -0.02 6296
## paternal_loss2025 -0.12 0.04 -0.19 -0.05 6398
## paternal_loss2530 -0.01 0.03 -0.07 0.06 6888
## paternal_loss3035 -0.03 0.03 -0.09 0.04 6319
## paternal_loss3540 -0.01 0.03 -0.07 0.06 6777
## paternal_loss4045 -0.01 0.04 -0.08 0.06 6360
## maternal_loss01 0.10 0.08 -0.06 0.25 6964
## maternal_loss15 -0.02 0.05 -0.11 0.07 6898
## maternal_loss510 0.07 0.04 -0.01 0.15 6798
## maternal_loss1015 0.03 0.04 -0.05 0.11 6059
## maternal_loss1520 0.00 0.04 -0.08 0.08 6999
## maternal_loss2025 0.00 0.04 -0.07 0.08 7138
## maternal_loss2530 -0.02 0.03 -0.09 0.04 7078
## maternal_loss3035 -0.05 0.03 -0.11 0.01 6812
## maternal_loss3540 -0.03 0.03 -0.09 0.02 6931
## maternal_loss4045 -0.03 0.03 -0.09 0.03 6981
## older_siblings1 0.02 0.03 -0.03 0.08 5936
## older_siblings2 -0.05 0.04 -0.12 0.02 5943
## older_siblings3 -0.07 0.05 -0.17 0.02 5885
## older_siblings4 -0.09 0.06 -0.21 0.03 5531
## older_siblings5P -0.10 0.08 -0.25 0.06 5554
## nr.siblings 0.01 0.01 0.00 0.02 5858
## last_born1 -0.04 0.02 -0.09 0.00 7200
## hu_Intercept -0.32 0.20 -0.71 0.07 6441
## hu_paternalage 0.26 0.14 -0.03 0.54 5609
## hu_birth_cohort1760M1765 -0.05 0.17 -0.38 0.29 6444
## hu_birth_cohort1765M1770 -0.31 0.15 -0.61 -0.02 6056
## hu_birth_cohort1770M1775 -0.07 0.15 -0.36 0.22 6451
## hu_birth_cohort1775M1780 -0.20 0.14 -0.47 0.08 5778
## hu_birth_cohort1780M1785 -0.27 0.15 -0.56 0.02 5681
## hu_birth_cohort1785M1790 -0.42 0.15 -0.71 -0.12 5891
## hu_birth_cohort1790M1795 -0.31 0.14 -0.58 -0.05 5570
## hu_birth_cohort1795M1800 -0.46 0.13 -0.71 -0.21 5679
## hu_birth_cohort1800M1805 -0.54 0.13 -0.78 -0.29 5475
## hu_birth_cohort1805M1810 -0.27 0.13 -0.52 -0.03 5868
## hu_birth_cohort1810M1815 -0.44 0.12 -0.68 -0.20 5451
## hu_birth_cohort1815M1820 -0.70 0.12 -0.93 -0.47 5462
## hu_birth_cohort1820M1825 -0.52 0.12 -0.76 -0.29 5472
## hu_birth_cohort1825M1830 -0.55 0.12 -0.78 -0.32 5631
## hu_birth_cohort1830M1835 -0.55 0.12 -0.79 -0.32 5467
## hu_male1 0.27 0.05 0.19 0.36 7200
## hu_maternalage.factor1420 0.24 0.23 -0.20 0.70 7200
## hu_maternalage.factor3550 0.13 0.07 -0.01 0.27 6260
## hu_paternalage.mean -0.18 0.15 -0.47 0.11 5627
## hu_paternal_loss01 0.58 0.18 0.23 0.93 6759
## hu_paternal_loss15 0.53 0.13 0.27 0.78 6115
## hu_paternal_loss510 0.19 0.11 -0.03 0.41 6933
## hu_paternal_loss1015 0.16 0.11 -0.05 0.36 6843
## hu_paternal_loss1520 0.10 0.10 -0.09 0.30 6789
## hu_paternal_loss2025 0.16 0.10 -0.04 0.35 6860
## hu_paternal_loss2530 0.06 0.09 -0.12 0.24 6632
## hu_paternal_loss3035 -0.02 0.09 -0.20 0.16 6543
## hu_paternal_loss3540 -0.01 0.09 -0.19 0.16 6848
## hu_paternal_loss4045 0.14 0.10 -0.05 0.34 6969
## hu_maternal_loss01 1.58 0.19 1.22 1.95 7200
## hu_maternal_loss15 0.59 0.12 0.35 0.82 6637
## hu_maternal_loss510 0.48 0.11 0.28 0.69 7032
## hu_maternal_loss1015 0.47 0.11 0.26 0.68 7200
## hu_maternal_loss1520 0.31 0.11 0.10 0.52 6745
## hu_maternal_loss2025 0.26 0.10 0.06 0.46 7003
## hu_maternal_loss2530 0.19 0.09 0.01 0.37 6168
## hu_maternal_loss3035 0.22 0.09 0.05 0.38 6961
## hu_maternal_loss3540 0.07 0.08 -0.09 0.23 7200
## hu_maternal_loss4045 0.28 0.08 0.11 0.45 7200
## hu_older_siblings1 -0.01 0.07 -0.15 0.14 6623
## hu_older_siblings2 -0.15 0.10 -0.34 0.04 6181
## hu_older_siblings3 -0.19 0.12 -0.44 0.05 5694
## hu_older_siblings4 -0.23 0.16 -0.53 0.08 5694
## hu_older_siblings5P -0.57 0.21 -0.97 -0.17 5652
## hu_nr.siblings 0.10 0.02 0.07 0.14 6153
## hu_last_born1 0.08 0.06 -0.04 0.21 7024
## Rhat
## Intercept 1
## paternalage 1
## birth_cohort1760M1765 1
## birth_cohort1765M1770 1
## birth_cohort1770M1775 1
## birth_cohort1775M1780 1
## birth_cohort1780M1785 1
## birth_cohort1785M1790 1
## birth_cohort1790M1795 1
## birth_cohort1795M1800 1
## birth_cohort1800M1805 1
## birth_cohort1805M1810 1
## birth_cohort1810M1815 1
## birth_cohort1815M1820 1
## birth_cohort1820M1825 1
## birth_cohort1825M1830 1
## birth_cohort1830M1835 1
## male1 1
## maternalage.factor1420 1
## maternalage.factor3550 1
## paternalage.mean 1
## paternal_loss01 1
## paternal_loss15 1
## paternal_loss510 1
## paternal_loss1015 1
## paternal_loss1520 1
## paternal_loss2025 1
## paternal_loss2530 1
## paternal_loss3035 1
## paternal_loss3540 1
## paternal_loss4045 1
## maternal_loss01 1
## maternal_loss15 1
## maternal_loss510 1
## maternal_loss1015 1
## maternal_loss1520 1
## maternal_loss2025 1
## maternal_loss2530 1
## maternal_loss3035 1
## maternal_loss3540 1
## maternal_loss4045 1
## older_siblings1 1
## older_siblings2 1
## older_siblings3 1
## older_siblings4 1
## older_siblings5P 1
## nr.siblings 1
## last_born1 1
## hu_Intercept 1
## hu_paternalage 1
## hu_birth_cohort1760M1765 1
## hu_birth_cohort1765M1770 1
## hu_birth_cohort1770M1775 1
## hu_birth_cohort1775M1780 1
## hu_birth_cohort1780M1785 1
## hu_birth_cohort1785M1790 1
## hu_birth_cohort1790M1795 1
## hu_birth_cohort1795M1800 1
## hu_birth_cohort1800M1805 1
## hu_birth_cohort1805M1810 1
## hu_birth_cohort1810M1815 1
## hu_birth_cohort1815M1820 1
## hu_birth_cohort1820M1825 1
## hu_birth_cohort1825M1830 1
## hu_birth_cohort1830M1835 1
## hu_male1 1
## hu_maternalage.factor1420 1
## hu_maternalage.factor3550 1
## hu_paternalage.mean 1
## hu_paternal_loss01 1
## hu_paternal_loss15 1
## hu_paternal_loss510 1
## hu_paternal_loss1015 1
## hu_paternal_loss1520 1
## hu_paternal_loss2025 1
## hu_paternal_loss2530 1
## hu_paternal_loss3035 1
## hu_paternal_loss3540 1
## hu_paternal_loss4045 1
## hu_maternal_loss01 1
## hu_maternal_loss15 1
## hu_maternal_loss510 1
## hu_maternal_loss1015 1
## hu_maternal_loss1520 1
## hu_maternal_loss2025 1
## hu_maternal_loss2530 1
## hu_maternal_loss3035 1
## hu_maternal_loss3540 1
## hu_maternal_loss4045 1
## hu_older_siblings1 1
## hu_older_siblings2 1
## hu_older_siblings3 1
## hu_older_siblings4 1
## hu_older_siblings5P 1
## hu_nr.siblings 1
## hu_last_born1 1
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.241 | 4.498 | 6.122 |
paternalage | 1.069 | 0.9607 | 1.187 |
birth_cohort1760M1765 | 0.9946 | 0.8784 | 1.128 |
birth_cohort1765M1770 | 0.8887 | 0.7923 | 0.994 |
birth_cohort1770M1775 | 0.8943 | 0.7976 | 1 |
birth_cohort1775M1780 | 0.9746 | 0.8739 | 1.087 |
birth_cohort1780M1785 | 0.8955 | 0.7998 | 1.001 |
birth_cohort1785M1790 | 0.9113 | 0.82 | 1.014 |
birth_cohort1790M1795 | 0.9281 | 0.8389 | 1.027 |
birth_cohort1795M1800 | 0.9016 | 0.8202 | 0.9939 |
birth_cohort1800M1805 | 0.8929 | 0.8119 | 0.9818 |
birth_cohort1805M1810 | 0.8729 | 0.7932 | 0.9588 |
birth_cohort1810M1815 | 0.9074 | 0.8275 | 0.9948 |
birth_cohort1815M1820 | 0.8685 | 0.7933 | 0.9491 |
birth_cohort1820M1825 | 0.8321 | 0.7611 | 0.9094 |
birth_cohort1825M1830 | 0.8106 | 0.7381 | 0.888 |
birth_cohort1830M1835 | 0.8339 | 0.7597 | 0.9164 |
male1 | 1.081 | 1.046 | 1.116 |
maternalage.factor1420 | 0.9458 | 0.7839 | 1.13 |
maternalage.factor3550 | 0.9972 | 0.9443 | 1.054 |
paternalage.mean | 0.9319 | 0.8369 | 1.039 |
paternal_loss01 | 0.8635 | 0.7439 | 0.9989 |
paternal_loss15 | 0.9705 | 0.8774 | 1.075 |
paternal_loss510 | 0.936 | 0.8585 | 1.021 |
paternal_loss1015 | 1.01 | 0.9327 | 1.094 |
paternal_loss1520 | 0.9116 | 0.8463 | 0.9835 |
paternal_loss2025 | 0.888 | 0.8263 | 0.9554 |
paternal_loss2530 | 0.9931 | 0.9288 | 1.064 |
paternal_loss3035 | 0.9744 | 0.9123 | 1.04 |
paternal_loss3540 | 0.9938 | 0.9313 | 1.058 |
paternal_loss4045 | 0.9921 | 0.9237 | 1.066 |
maternal_loss01 | 1.105 | 0.9442 | 1.286 |
maternal_loss15 | 0.9789 | 0.8915 | 1.071 |
maternal_loss510 | 1.071 | 0.99 | 1.16 |
maternal_loss1015 | 1.027 | 0.9465 | 1.112 |
maternal_loss1520 | 1.003 | 0.926 | 1.087 |
maternal_loss2025 | 1.003 | 0.9293 | 1.085 |
maternal_loss2530 | 0.9777 | 0.9144 | 1.045 |
maternal_loss3035 | 0.9504 | 0.8915 | 1.013 |
maternal_loss3540 | 0.9669 | 0.912 | 1.023 |
maternal_loss4045 | 0.9711 | 0.9125 | 1.033 |
older_siblings1 | 1.025 | 0.9722 | 1.083 |
older_siblings2 | 0.9535 | 0.8866 | 1.024 |
older_siblings3 | 0.9288 | 0.8479 | 1.019 |
older_siblings4 | 0.9117 | 0.8113 | 1.026 |
older_siblings5P | 0.9082 | 0.7798 | 1.061 |
nr.siblings | 1.01 | 0.9955 | 1.024 |
last_born1 | 0.9577 | 0.913 | 1.004 |
hu_Intercept | 0.7266 | 0.4938 | 1.074 |
hu_paternalage | 1.295 | 0.975 | 1.715 |
hu_birth_cohort1760M1765 | 0.9496 | 0.6837 | 1.331 |
hu_birth_cohort1765M1770 | 0.7308 | 0.542 | 0.9769 |
hu_birth_cohort1770M1775 | 0.9347 | 0.7009 | 1.245 |
hu_birth_cohort1775M1780 | 0.8204 | 0.6225 | 1.081 |
hu_birth_cohort1780M1785 | 0.7598 | 0.5726 | 1.017 |
hu_birth_cohort1785M1790 | 0.6556 | 0.4921 | 0.8838 |
hu_birth_cohort1790M1795 | 0.731 | 0.5579 | 0.9521 |
hu_birth_cohort1795M1800 | 0.6325 | 0.4902 | 0.8137 |
hu_birth_cohort1800M1805 | 0.5853 | 0.4569 | 0.7497 |
hu_birth_cohort1805M1810 | 0.7644 | 0.5937 | 0.9752 |
hu_birth_cohort1810M1815 | 0.6438 | 0.5054 | 0.8209 |
hu_birth_cohort1815M1820 | 0.4943 | 0.3937 | 0.6221 |
hu_birth_cohort1820M1825 | 0.5918 | 0.4678 | 0.7494 |
hu_birth_cohort1825M1830 | 0.579 | 0.4578 | 0.7294 |
hu_birth_cohort1830M1835 | 0.5744 | 0.4521 | 0.7297 |
hu_male1 | 1.315 | 1.203 | 1.438 |
hu_maternalage.factor1420 | 1.271 | 0.8165 | 2.006 |
hu_maternalage.factor3550 | 1.142 | 0.9916 | 1.313 |
hu_paternalage.mean | 0.8357 | 0.6237 | 1.12 |
hu_paternal_loss01 | 1.784 | 1.254 | 2.547 |
hu_paternal_loss15 | 1.691 | 1.31 | 2.184 |
hu_paternal_loss510 | 1.211 | 0.9682 | 1.503 |
hu_paternal_loss1015 | 1.17 | 0.9529 | 1.438 |
hu_paternal_loss1520 | 1.11 | 0.91 | 1.35 |
hu_paternal_loss2025 | 1.168 | 0.9629 | 1.418 |
hu_paternal_loss2530 | 1.062 | 0.8833 | 1.273 |
hu_paternal_loss3035 | 0.9789 | 0.8201 | 1.172 |
hu_paternal_loss3540 | 0.9856 | 0.8275 | 1.17 |
hu_paternal_loss4045 | 1.153 | 0.9487 | 1.398 |
hu_maternal_loss01 | 4.844 | 3.381 | 7.038 |
hu_maternal_loss15 | 1.798 | 1.423 | 2.269 |
hu_maternal_loss510 | 1.623 | 1.318 | 1.986 |
hu_maternal_loss1015 | 1.597 | 1.299 | 1.965 |
hu_maternal_loss1520 | 1.363 | 1.105 | 1.685 |
hu_maternal_loss2025 | 1.299 | 1.066 | 1.589 |
hu_maternal_loss2530 | 1.208 | 1.007 | 1.454 |
hu_maternal_loss3035 | 1.244 | 1.055 | 1.468 |
hu_maternal_loss3540 | 1.07 | 0.9152 | 1.254 |
hu_maternal_loss4045 | 1.321 | 1.121 | 1.562 |
hu_older_siblings1 | 0.991 | 0.8579 | 1.145 |
hu_older_siblings2 | 0.858 | 0.7107 | 1.037 |
hu_older_siblings3 | 0.8247 | 0.6471 | 1.053 |
hu_older_siblings4 | 0.7973 | 0.5869 | 1.079 |
hu_older_siblings5P | 0.5668 | 0.3796 | 0.8449 |
hu_nr.siblings | 1.109 | 1.07 | 1.15 |
hu_last_born1 | 1.085 | 0.9581 | 1.23 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.28 | [1.92;2.67] | [2.04;2.53] |
estimate father 35y | 2.09 | [1.65;2.59] | [1.79;2.4] |
percentage change | -8.41 | [-25.14;11.23] | [-19.31;3.87] |
OR/IRR | 1.07 | [0.96;1.19] | [1;1.15] |
OR hurdle | 1.3 | [0.98;1.71] | [1.08;1.55] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r11_separate_random_effects_for_parents.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
It need not be the case that paternal age has the same effect on male and female children. For example, male children inherit only the small Y chromosome from the father, but female children inherit the larger X chromosome, so that paternal age predicts X-chromosomal de novo mutations in females but not in males (Francioli et al., 2016). At the same time, the autism literature suggests that males are less robust to heritable and de novo autism risk variants and that these effects are not simply due to having only one X chromosome (Werling & Geschwind, 2015). Here we let a dummy variable for being male moderate the paternal age effect.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage * male + birth_cohort + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + male + birth_cohort + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents) + paternalage:male
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.23 0.01 0.20 0.25 1029 1.00
## sd(hu_Intercept) 0.48 0.05 0.38 0.57 728 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.66 0.09 1.49 1.83 1372
## paternalage 0.07 0.06 -0.04 0.18 690
## male1 0.07 0.08 -0.09 0.22 2537
## birth_cohort1760M1765 0.00 0.06 -0.12 0.12 1423
## birth_cohort1765M1770 -0.12 0.06 -0.23 0.00 983
## birth_cohort1770M1775 -0.11 0.06 -0.23 0.00 916
## birth_cohort1775M1780 -0.03 0.06 -0.13 0.09 936
## birth_cohort1780M1785 -0.11 0.06 -0.23 0.00 933
## birth_cohort1785M1790 -0.10 0.05 -0.20 0.01 856
## birth_cohort1790M1795 -0.08 0.05 -0.18 0.03 778
## birth_cohort1795M1800 -0.11 0.05 -0.20 -0.01 761
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.02 708
## birth_cohort1805M1810 -0.14 0.05 -0.23 -0.04 786
## birth_cohort1810M1815 -0.10 0.05 -0.19 0.00 675
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 697
## birth_cohort1820M1825 -0.19 0.05 -0.28 -0.10 674
## birth_cohort1825M1830 -0.21 0.05 -0.30 -0.12 678
## birth_cohort1830M1835 -0.18 0.05 -0.28 -0.09 720
## maternalage.factor1420 -0.05 0.09 -0.23 0.13 3000
## maternalage.factor3550 0.00 0.03 -0.06 0.05 3000
## paternalage.mean -0.07 0.06 -0.18 0.04 735
## paternal_loss01 -0.15 0.08 -0.30 0.00 3000
## paternal_loss15 -0.04 0.05 -0.14 0.06 2087
## paternal_loss510 -0.07 0.04 -0.15 0.02 1729
## paternal_loss1015 0.01 0.04 -0.07 0.09 1755
## paternal_loss1520 -0.09 0.04 -0.17 -0.02 1714
## paternal_loss2025 -0.12 0.04 -0.19 -0.05 1697
## paternal_loss2530 -0.01 0.03 -0.08 0.05 1607
## paternal_loss3035 -0.03 0.03 -0.09 0.03 1484
## paternal_loss3540 -0.01 0.03 -0.07 0.06 1899
## paternal_loss4045 -0.01 0.04 -0.08 0.06 3000
## maternal_loss01 0.10 0.08 -0.06 0.25 3000
## maternal_loss15 -0.02 0.05 -0.11 0.07 3000
## maternal_loss510 0.07 0.04 -0.01 0.15 3000
## maternal_loss1015 0.03 0.04 -0.05 0.11 3000
## maternal_loss1520 0.00 0.04 -0.08 0.09 3000
## maternal_loss2025 0.00 0.04 -0.07 0.08 3000
## maternal_loss2530 -0.02 0.04 -0.09 0.04 3000
## maternal_loss3035 -0.05 0.03 -0.12 0.01 3000
## maternal_loss3540 -0.03 0.03 -0.09 0.02 3000
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## older_siblings1 0.03 0.03 -0.03 0.08 1749
## older_siblings2 -0.05 0.04 -0.12 0.02 971
## older_siblings3 -0.07 0.05 -0.17 0.02 761
## older_siblings4 -0.09 0.06 -0.21 0.03 698
## older_siblings5P -0.10 0.08 -0.26 0.06 663
## nr.siblings 0.01 0.01 0.00 0.02 764
## last_born1 -0.04 0.02 -0.09 0.00 3000
## paternalage:male1 0.00 0.02 -0.04 0.05 2481
## hu_Intercept -0.35 0.22 -0.78 0.10 1351
## hu_paternalage 0.27 0.14 -0.01 0.55 783
## hu_male1 0.30 0.21 -0.10 0.73 2166
## hu_birth_cohort1760M1765 -0.05 0.17 -0.37 0.28 3000
## hu_birth_cohort1765M1770 -0.31 0.14 -0.58 -0.03 3000
## hu_birth_cohort1770M1775 -0.07 0.14 -0.35 0.21 1453
## hu_birth_cohort1775M1780 -0.20 0.14 -0.47 0.07 1401
## hu_birth_cohort1780M1785 -0.28 0.14 -0.56 0.01 3000
## hu_birth_cohort1785M1790 -0.42 0.14 -0.70 -0.14 1593
## hu_birth_cohort1790M1795 -0.31 0.13 -0.58 -0.06 1100
## hu_birth_cohort1795M1800 -0.46 0.12 -0.69 -0.22 1006
## hu_birth_cohort1800M1805 -0.53 0.12 -0.78 -0.30 594
## hu_birth_cohort1805M1810 -0.27 0.12 -0.51 -0.03 767
## hu_birth_cohort1810M1815 -0.44 0.12 -0.67 -0.20 1083
## hu_birth_cohort1815M1820 -0.70 0.11 -0.93 -0.47 886
## hu_birth_cohort1820M1825 -0.52 0.11 -0.75 -0.30 878
## hu_birth_cohort1825M1830 -0.54 0.11 -0.77 -0.32 849
## hu_birth_cohort1830M1835 -0.55 0.12 -0.78 -0.32 580
## hu_maternalage.factor1420 0.24 0.23 -0.21 0.70 3000
## hu_maternalage.factor3550 0.13 0.07 -0.01 0.27 3000
## hu_paternalage.mean -0.18 0.14 -0.46 0.10 854
## hu_paternal_loss01 0.58 0.19 0.22 0.94 3000
## hu_paternal_loss15 0.53 0.13 0.28 0.80 3000
## hu_paternal_loss510 0.20 0.11 -0.03 0.42 1793
## hu_paternal_loss1015 0.16 0.11 -0.05 0.36 1579
## hu_paternal_loss1520 0.10 0.10 -0.10 0.30 1473
## hu_paternal_loss2025 0.15 0.10 -0.03 0.34 1411
## hu_paternal_loss2530 0.06 0.09 -0.12 0.24 1425
## hu_paternal_loss3035 -0.02 0.09 -0.20 0.15 1352
## hu_paternal_loss3540 -0.01 0.09 -0.19 0.16 1563
## hu_paternal_loss4045 0.14 0.10 -0.07 0.34 3000
## hu_maternal_loss01 1.58 0.18 1.23 1.94 3000
## hu_maternal_loss15 0.59 0.12 0.36 0.82 3000
## hu_maternal_loss510 0.48 0.11 0.28 0.69 3000
## hu_maternal_loss1015 0.47 0.10 0.27 0.68 3000
## hu_maternal_loss1520 0.31 0.10 0.11 0.52 3000
## hu_maternal_loss2025 0.26 0.10 0.07 0.46 3000
## hu_maternal_loss2530 0.19 0.09 0.00 0.37 3000
## hu_maternal_loss3035 0.21 0.09 0.04 0.38 3000
## hu_maternal_loss3540 0.07 0.08 -0.09 0.22 3000
## hu_maternal_loss4045 0.28 0.08 0.11 0.45 3000
## hu_older_siblings1 -0.01 0.07 -0.16 0.13 1597
## hu_older_siblings2 -0.16 0.10 -0.35 0.02 955
## hu_older_siblings3 -0.20 0.13 -0.43 0.05 817
## hu_older_siblings4 -0.23 0.15 -0.53 0.06 775
## hu_older_siblings5P -0.58 0.20 -0.98 -0.19 755
## hu_nr.siblings 0.10 0.02 0.07 0.14 1054
## hu_last_born1 0.08 0.06 -0.04 0.20 3000
## hu_paternalage:male1 -0.01 0.06 -0.13 0.11 2259
## Rhat
## Intercept 1.00
## paternalage 1.01
## male1 1.00
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.00
## birth_cohort1770M1775 1.01
## birth_cohort1775M1780 1.00
## birth_cohort1780M1785 1.00
## birth_cohort1785M1790 1.01
## birth_cohort1790M1795 1.01
## birth_cohort1795M1800 1.01
## birth_cohort1800M1805 1.01
## birth_cohort1805M1810 1.00
## birth_cohort1810M1815 1.00
## birth_cohort1815M1820 1.00
## birth_cohort1820M1825 1.01
## birth_cohort1825M1830 1.01
## birth_cohort1830M1835 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.00
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.00
## older_siblings3 1.00
## older_siblings4 1.01
## older_siblings5P 1.01
## nr.siblings 1.01
## last_born1 1.00
## paternalage:male1 1.00
## hu_Intercept 1.00
## hu_paternalage 1.01
## hu_male1 1.00
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.00
## hu_birth_cohort1775M1780 1.00
## hu_birth_cohort1780M1785 1.00
## hu_birth_cohort1785M1790 1.00
## hu_birth_cohort1790M1795 1.00
## hu_birth_cohort1795M1800 1.00
## hu_birth_cohort1800M1805 1.00
## hu_birth_cohort1805M1810 1.00
## hu_birth_cohort1810M1815 1.00
## hu_birth_cohort1815M1820 1.00
## hu_birth_cohort1820M1825 1.00
## hu_birth_cohort1825M1830 1.00
## hu_birth_cohort1830M1835 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.01
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.01
## hu_older_siblings3 1.02
## hu_older_siblings4 1.01
## hu_older_siblings5P 1.02
## hu_nr.siblings 1.01
## hu_last_born1 1.00
## hu_paternalage:male1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.234 | 4.431 | 6.222 |
paternalage | 1.069 | 0.959 | 1.199 |
male1 | 1.075 | 0.9174 | 1.252 |
birth_cohort1760M1765 | 0.9968 | 0.8842 | 1.124 |
birth_cohort1765M1770 | 0.8874 | 0.7925 | 0.996 |
birth_cohort1770M1775 | 0.8916 | 0.7966 | 0.9962 |
birth_cohort1775M1780 | 0.9742 | 0.876 | 1.091 |
birth_cohort1780M1785 | 0.8962 | 0.7965 | 1.002 |
birth_cohort1785M1790 | 0.9078 | 0.8172 | 1.009 |
birth_cohort1790M1795 | 0.9262 | 0.8339 | 1.028 |
birth_cohort1795M1800 | 0.8999 | 0.8172 | 0.9893 |
birth_cohort1800M1805 | 0.8919 | 0.8112 | 0.9781 |
birth_cohort1805M1810 | 0.873 | 0.7922 | 0.9616 |
birth_cohort1810M1815 | 0.9072 | 0.8266 | 0.9967 |
birth_cohort1815M1820 | 0.8679 | 0.7929 | 0.95 |
birth_cohort1820M1825 | 0.8305 | 0.7575 | 0.9088 |
birth_cohort1825M1830 | 0.8102 | 0.7378 | 0.8859 |
birth_cohort1830M1835 | 0.8332 | 0.7584 | 0.9129 |
maternalage.factor1420 | 0.9476 | 0.795 | 1.134 |
maternalage.factor3550 | 0.9971 | 0.9446 | 1.054 |
paternalage.mean | 0.9327 | 0.8316 | 1.039 |
paternal_loss01 | 0.8576 | 0.7407 | 1.001 |
paternal_loss15 | 0.9638 | 0.8701 | 1.063 |
paternal_loss510 | 0.9351 | 0.8609 | 1.017 |
paternal_loss1015 | 1.007 | 0.9296 | 1.089 |
paternal_loss1520 | 0.9116 | 0.8438 | 0.9804 |
paternal_loss2025 | 0.8878 | 0.8257 | 0.953 |
paternal_loss2530 | 0.991 | 0.9271 | 1.056 |
paternal_loss3035 | 0.9723 | 0.9129 | 1.035 |
paternal_loss3540 | 0.9913 | 0.9326 | 1.058 |
paternal_loss4045 | 0.9907 | 0.921 | 1.063 |
maternal_loss01 | 1.105 | 0.9397 | 1.287 |
maternal_loss15 | 0.9804 | 0.8945 | 1.072 |
maternal_loss510 | 1.073 | 0.9923 | 1.159 |
maternal_loss1015 | 1.027 | 0.9486 | 1.111 |
maternal_loss1520 | 1.003 | 0.9255 | 1.089 |
maternal_loss2025 | 1.005 | 0.9329 | 1.081 |
maternal_loss2530 | 0.9785 | 0.9118 | 1.046 |
maternal_loss3035 | 0.9503 | 0.89 | 1.01 |
maternal_loss3540 | 0.9678 | 0.9126 | 1.021 |
maternal_loss4045 | 0.972 | 0.9152 | 1.034 |
older_siblings1 | 1.026 | 0.9732 | 1.084 |
older_siblings2 | 0.953 | 0.8855 | 1.024 |
older_siblings3 | 0.9292 | 0.8455 | 1.02 |
older_siblings4 | 0.9113 | 0.8073 | 1.026 |
older_siblings5P | 0.9078 | 0.7743 | 1.063 |
nr.siblings | 1.01 | 0.9953 | 1.024 |
last_born1 | 0.9572 | 0.9146 | 1.002 |
paternalage:male1 | 1.002 | 0.9593 | 1.048 |
hu_Intercept | 0.7072 | 0.4605 | 1.103 |
hu_paternalage | 1.307 | 0.9891 | 1.731 |
hu_male1 | 1.354 | 0.909 | 2.069 |
hu_birth_cohort1760M1765 | 0.9518 | 0.6918 | 1.32 |
hu_birth_cohort1765M1770 | 0.7366 | 0.5582 | 0.969 |
hu_birth_cohort1770M1775 | 0.9358 | 0.7081 | 1.235 |
hu_birth_cohort1775M1780 | 0.8219 | 0.6275 | 1.071 |
hu_birth_cohort1780M1785 | 0.758 | 0.5704 | 1.005 |
hu_birth_cohort1785M1790 | 0.6589 | 0.4962 | 0.8668 |
hu_birth_cohort1790M1795 | 0.7329 | 0.5612 | 0.9434 |
hu_birth_cohort1795M1800 | 0.634 | 0.5001 | 0.8051 |
hu_birth_cohort1800M1805 | 0.5858 | 0.4586 | 0.7379 |
hu_birth_cohort1805M1810 | 0.7657 | 0.6016 | 0.9733 |
hu_birth_cohort1810M1815 | 0.6452 | 0.5101 | 0.8151 |
hu_birth_cohort1815M1820 | 0.4959 | 0.3931 | 0.6254 |
hu_birth_cohort1820M1825 | 0.5935 | 0.4735 | 0.7408 |
hu_birth_cohort1825M1830 | 0.5812 | 0.4626 | 0.7292 |
hu_birth_cohort1830M1835 | 0.5762 | 0.4563 | 0.7243 |
hu_maternalage.factor1420 | 1.269 | 0.8077 | 2.004 |
hu_maternalage.factor3550 | 1.142 | 0.9889 | 1.312 |
hu_paternalage.mean | 0.8342 | 0.6308 | 1.101 |
hu_paternal_loss01 | 1.785 | 1.241 | 2.557 |
hu_paternal_loss15 | 1.697 | 1.323 | 2.221 |
hu_paternal_loss510 | 1.217 | 0.9739 | 1.519 |
hu_paternal_loss1015 | 1.174 | 0.956 | 1.439 |
hu_paternal_loss1520 | 1.108 | 0.9071 | 1.347 |
hu_paternal_loss2025 | 1.166 | 0.967 | 1.409 |
hu_paternal_loss2530 | 1.061 | 0.8888 | 1.274 |
hu_paternal_loss3035 | 0.9777 | 0.817 | 1.164 |
hu_paternal_loss3540 | 0.9865 | 0.8257 | 1.176 |
hu_paternal_loss4045 | 1.153 | 0.9351 | 1.407 |
hu_maternal_loss01 | 4.863 | 3.423 | 6.949 |
hu_maternal_loss15 | 1.8 | 1.433 | 2.277 |
hu_maternal_loss510 | 1.624 | 1.32 | 1.991 |
hu_maternal_loss1015 | 1.606 | 1.312 | 1.968 |
hu_maternal_loss1520 | 1.369 | 1.117 | 1.675 |
hu_maternal_loss2025 | 1.299 | 1.073 | 1.583 |
hu_maternal_loss2530 | 1.21 | 1.004 | 1.453 |
hu_maternal_loss3035 | 1.239 | 1.044 | 1.469 |
hu_maternal_loss3540 | 1.07 | 0.9124 | 1.251 |
hu_maternal_loss4045 | 1.319 | 1.121 | 1.564 |
hu_older_siblings1 | 0.9858 | 0.848 | 1.137 |
hu_older_siblings2 | 0.8538 | 0.7055 | 1.024 |
hu_older_siblings3 | 0.8215 | 0.6479 | 1.053 |
hu_older_siblings4 | 0.7911 | 0.5892 | 1.066 |
hu_older_siblings5P | 0.5611 | 0.3767 | 0.8303 |
hu_nr.siblings | 1.11 | 1.071 | 1.15 |
hu_last_born1 | 1.084 | 0.9612 | 1.226 |
hu_paternalage:male1 | 0.9917 | 0.8812 | 1.112 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.29 | [1.94;2.7] | [2.05;2.54] |
estimate father 35y | 2.09 | [1.65;2.59] | [1.8;2.41] |
percentage change | -9.05 | [-25.49;11.82] | [-20.29;4.37] |
OR/IRR | 1.07 | [0.96;1.2] | [0.99;1.15] |
OR hurdle | 1.3 | [0.99;1.73] | [1.09;1.57] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r12_sex_moderation.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
We already control for the average paternal age at which the children in a family were born. The mean is a more complete summary of the reproductive timing of the father than the age at first birth. However, far more literature has examined age at first birth and it has the advantage of never being censored (although we of course try to rule out censoring by choosing appropriate subsets). Therefore, we added age at first birth as a covariate in this model.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage_at_1st_sib + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage_at_1st_sib + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 1500; warmup = 1000; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.22 0.01 0.20 0.25 860 1
## sd(hu_Intercept) 0.47 0.04 0.39 0.56 818 1
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.66 0.08 1.50 1.82 1348
## paternalage 0.07 0.06 -0.04 0.18 708
## birth_cohort1760M1765 0.00 0.06 -0.12 0.12 1182
## birth_cohort1765M1770 -0.12 0.06 -0.23 0.00 1007
## birth_cohort1770M1775 -0.11 0.06 -0.23 0.00 869
## birth_cohort1775M1780 -0.02 0.06 -0.14 0.09 786
## birth_cohort1780M1785 -0.11 0.06 -0.23 0.01 881
## birth_cohort1785M1790 -0.09 0.06 -0.21 0.01 739
## birth_cohort1790M1795 -0.07 0.05 -0.18 0.03 680
## birth_cohort1795M1800 -0.10 0.05 -0.20 0.00 674
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.01 613
## birth_cohort1805M1810 -0.13 0.05 -0.23 -0.04 636
## birth_cohort1810M1815 -0.10 0.05 -0.19 0.00 610
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 538
## birth_cohort1820M1825 -0.18 0.05 -0.27 -0.09 551
## birth_cohort1825M1830 -0.21 0.05 -0.30 -0.11 565
## birth_cohort1830M1835 -0.18 0.05 -0.28 -0.09 558
## male1 0.08 0.02 0.05 0.11 3000
## maternalage.factor1420 -0.05 0.09 -0.24 0.12 3000
## maternalage.factor3550 0.00 0.03 -0.06 0.05 3000
## paternalage_at_1st_sib -0.01 0.03 -0.08 0.05 3000
## paternalage.mean -0.06 0.06 -0.18 0.06 793
## paternal_loss01 -0.15 0.07 -0.30 -0.01 3000
## paternal_loss15 -0.04 0.05 -0.14 0.06 1831
## paternal_loss510 -0.07 0.04 -0.16 0.02 1573
## paternal_loss1015 0.01 0.04 -0.07 0.09 1428
## paternal_loss1520 -0.09 0.04 -0.17 -0.02 1574
## paternal_loss2025 -0.12 0.04 -0.19 -0.05 1466
## paternal_loss2530 -0.01 0.03 -0.08 0.06 1382
## paternal_loss3035 -0.03 0.03 -0.10 0.04 1417
## paternal_loss3540 -0.01 0.03 -0.07 0.05 1473
## paternal_loss4045 -0.01 0.04 -0.08 0.06 3000
## maternal_loss01 0.10 0.08 -0.06 0.26 3000
## maternal_loss15 -0.02 0.05 -0.11 0.08 3000
## maternal_loss510 0.07 0.04 -0.01 0.15 3000
## maternal_loss1015 0.03 0.04 -0.05 0.11 3000
## maternal_loss1520 0.00 0.04 -0.07 0.08 3000
## maternal_loss2025 0.00 0.04 -0.07 0.08 2007
## maternal_loss2530 -0.02 0.03 -0.09 0.05 1932
## maternal_loss3035 -0.05 0.03 -0.12 0.01 1904
## maternal_loss3540 -0.03 0.03 -0.09 0.03 3000
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## older_siblings1 0.03 0.03 -0.03 0.08 1335
## older_siblings2 -0.05 0.04 -0.12 0.02 820
## older_siblings3 -0.07 0.05 -0.17 0.02 689
## older_siblings4 -0.09 0.06 -0.22 0.02 739
## older_siblings5P -0.10 0.08 -0.26 0.06 648
## nr.siblings 0.01 0.01 -0.01 0.02 764
## last_born1 -0.04 0.02 -0.09 0.00 3000
## hu_Intercept -0.38 0.20 -0.78 0.01 1116
## hu_paternalage 0.26 0.15 -0.03 0.53 796
## hu_birth_cohort1760M1765 -0.05 0.17 -0.38 0.29 3000
## hu_birth_cohort1765M1770 -0.31 0.15 -0.60 -0.03 1122
## hu_birth_cohort1770M1775 -0.06 0.15 -0.36 0.22 1030
## hu_birth_cohort1775M1780 -0.20 0.14 -0.48 0.09 1005
## hu_birth_cohort1780M1785 -0.28 0.14 -0.56 0.00 1006
## hu_birth_cohort1785M1790 -0.42 0.14 -0.70 -0.14 961
## hu_birth_cohort1790M1795 -0.32 0.13 -0.58 -0.06 798
## hu_birth_cohort1795M1800 -0.47 0.13 -0.72 -0.21 755
## hu_birth_cohort1800M1805 -0.54 0.13 -0.80 -0.29 837
## hu_birth_cohort1805M1810 -0.27 0.13 -0.52 -0.02 813
## hu_birth_cohort1810M1815 -0.44 0.12 -0.67 -0.21 808
## hu_birth_cohort1815M1820 -0.71 0.12 -0.94 -0.48 654
## hu_birth_cohort1820M1825 -0.53 0.12 -0.76 -0.30 697
## hu_birth_cohort1825M1830 -0.55 0.12 -0.78 -0.32 634
## hu_birth_cohort1830M1835 -0.56 0.12 -0.79 -0.32 699
## hu_male1 0.28 0.05 0.18 0.36 3000
## hu_maternalage.factor1420 0.24 0.23 -0.21 0.69 3000
## hu_maternalage.factor3550 0.14 0.07 -0.01 0.28 3000
## hu_paternalage_at_1st_sib 0.08 0.08 -0.07 0.23 3000
## hu_paternalage.mean -0.23 0.15 -0.53 0.07 849
## hu_paternal_loss01 0.57 0.19 0.20 0.95 3000
## hu_paternal_loss15 0.53 0.13 0.27 0.79 3000
## hu_paternal_loss510 0.20 0.11 -0.02 0.42 1537
## hu_paternal_loss1015 0.16 0.11 -0.06 0.37 1377
## hu_paternal_loss1520 0.10 0.10 -0.10 0.29 1334
## hu_paternal_loss2025 0.15 0.10 -0.04 0.35 1349
## hu_paternal_loss2530 0.06 0.09 -0.12 0.24 1274
## hu_paternal_loss3035 -0.03 0.09 -0.21 0.15 1360
## hu_paternal_loss3540 -0.02 0.09 -0.20 0.16 1428
## hu_paternal_loss4045 0.14 0.10 -0.05 0.33 3000
## hu_maternal_loss01 1.57 0.19 1.22 1.94 3000
## hu_maternal_loss15 0.58 0.12 0.35 0.82 3000
## hu_maternal_loss510 0.48 0.11 0.27 0.69 3000
## hu_maternal_loss1015 0.47 0.11 0.24 0.68 3000
## hu_maternal_loss1520 0.31 0.11 0.10 0.52 3000
## hu_maternal_loss2025 0.26 0.10 0.06 0.46 3000
## hu_maternal_loss2530 0.19 0.09 0.00 0.37 3000
## hu_maternal_loss3035 0.22 0.09 0.05 0.39 3000
## hu_maternal_loss3540 0.07 0.08 -0.09 0.23 3000
## hu_maternal_loss4045 0.28 0.09 0.10 0.44 3000
## hu_older_siblings1 -0.01 0.08 -0.16 0.14 3000
## hu_older_siblings2 -0.15 0.10 -0.34 0.04 963
## hu_older_siblings3 -0.19 0.12 -0.43 0.05 904
## hu_older_siblings4 -0.23 0.16 -0.53 0.07 829
## hu_older_siblings5P -0.57 0.21 -0.96 -0.17 773
## hu_nr.siblings 0.11 0.02 0.07 0.15 1000
## hu_last_born1 0.08 0.06 -0.04 0.20 3000
## Rhat
## Intercept 1.00
## paternalage 1.01
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.00
## birth_cohort1770M1775 1.00
## birth_cohort1775M1780 1.01
## birth_cohort1780M1785 1.00
## birth_cohort1785M1790 1.00
## birth_cohort1790M1795 1.01
## birth_cohort1795M1800 1.01
## birth_cohort1800M1805 1.01
## birth_cohort1805M1810 1.00
## birth_cohort1810M1815 1.01
## birth_cohort1815M1820 1.01
## birth_cohort1820M1825 1.01
## birth_cohort1825M1830 1.01
## birth_cohort1830M1835 1.01
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage_at_1st_sib 1.00
## paternalage.mean 1.00
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.00
## older_siblings3 1.01
## older_siblings4 1.00
## older_siblings5P 1.01
## nr.siblings 1.01
## last_born1 1.00
## hu_Intercept 1.00
## hu_paternalage 1.01
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.00
## hu_birth_cohort1775M1780 1.00
## hu_birth_cohort1780M1785 1.00
## hu_birth_cohort1785M1790 1.01
## hu_birth_cohort1790M1795 1.00
## hu_birth_cohort1795M1800 1.00
## hu_birth_cohort1800M1805 1.01
## hu_birth_cohort1805M1810 1.00
## hu_birth_cohort1810M1815 1.01
## hu_birth_cohort1815M1820 1.01
## hu_birth_cohort1820M1825 1.00
## hu_birth_cohort1825M1830 1.00
## hu_birth_cohort1830M1835 1.00
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage_at_1st_sib 1.00
## hu_paternalage.mean 1.01
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.00
## hu_older_siblings3 1.01
## hu_older_siblings4 1.01
## hu_older_siblings5P 1.01
## hu_nr.siblings 1.01
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.283 | 4.478 | 6.171 |
paternalage | 1.072 | 0.9582 | 1.192 |
birth_cohort1760M1765 | 0.998 | 0.8843 | 1.129 |
birth_cohort1765M1770 | 0.8891 | 0.795 | 0.9968 |
birth_cohort1770M1775 | 0.8928 | 0.7962 | 1.005 |
birth_cohort1775M1780 | 0.976 | 0.8703 | 1.091 |
birth_cohort1780M1785 | 0.8976 | 0.7978 | 1.007 |
birth_cohort1785M1790 | 0.9105 | 0.8125 | 1.014 |
birth_cohort1790M1795 | 0.9287 | 0.8344 | 1.029 |
birth_cohort1795M1800 | 0.9021 | 0.8158 | 0.9951 |
birth_cohort1800M1805 | 0.8946 | 0.812 | 0.9857 |
birth_cohort1805M1810 | 0.874 | 0.7924 | 0.963 |
birth_cohort1810M1815 | 0.9085 | 0.8294 | 0.998 |
birth_cohort1815M1820 | 0.869 | 0.7942 | 0.9543 |
birth_cohort1820M1825 | 0.8322 | 0.7601 | 0.9131 |
birth_cohort1825M1830 | 0.8121 | 0.7416 | 0.8947 |
birth_cohort1830M1835 | 0.8345 | 0.7562 | 0.9153 |
male1 | 1.081 | 1.046 | 1.116 |
maternalage.factor1420 | 0.9497 | 0.7879 | 1.132 |
maternalage.factor3550 | 0.9957 | 0.9444 | 1.051 |
paternalage_at_1st_sib | 0.9866 | 0.9243 | 1.049 |
paternalage.mean | 0.9388 | 0.8366 | 1.062 |
paternal_loss01 | 0.8576 | 0.7379 | 0.9888 |
paternal_loss15 | 0.9628 | 0.8689 | 1.063 |
paternal_loss510 | 0.9325 | 0.8557 | 1.017 |
paternal_loss1015 | 1.006 | 0.9302 | 1.09 |
paternal_loss1520 | 0.9103 | 0.8426 | 0.9779 |
paternal_loss2025 | 0.8867 | 0.825 | 0.9553 |
paternal_loss2530 | 0.9905 | 0.926 | 1.06 |
paternal_loss3035 | 0.9723 | 0.9085 | 1.038 |
paternal_loss3540 | 0.9909 | 0.9312 | 1.055 |
paternal_loss4045 | 0.9901 | 0.92 | 1.064 |
maternal_loss01 | 1.109 | 0.9384 | 1.297 |
maternal_loss15 | 0.9818 | 0.8933 | 1.078 |
maternal_loss510 | 1.074 | 0.9949 | 1.16 |
maternal_loss1015 | 1.027 | 0.9493 | 1.112 |
maternal_loss1520 | 1.003 | 0.9282 | 1.084 |
maternal_loss2025 | 1.003 | 0.9325 | 1.081 |
maternal_loss2530 | 0.9781 | 0.9147 | 1.046 |
maternal_loss3035 | 0.9493 | 0.8895 | 1.013 |
maternal_loss3540 | 0.9665 | 0.9099 | 1.028 |
maternal_loss4045 | 0.9716 | 0.9138 | 1.032 |
older_siblings1 | 1.026 | 0.9732 | 1.081 |
older_siblings2 | 0.9525 | 0.8853 | 1.025 |
older_siblings3 | 0.9279 | 0.8439 | 1.023 |
older_siblings4 | 0.9096 | 0.8038 | 1.023 |
older_siblings5P | 0.9068 | 0.7727 | 1.066 |
nr.siblings | 1.009 | 0.994 | 1.025 |
last_born1 | 0.9574 | 0.914 | 1.005 |
hu_Intercept | 0.6826 | 0.4576 | 1.013 |
hu_paternalage | 1.295 | 0.9666 | 1.699 |
hu_birth_cohort1760M1765 | 0.9509 | 0.6857 | 1.334 |
hu_birth_cohort1765M1770 | 0.7322 | 0.5465 | 0.9731 |
hu_birth_cohort1770M1775 | 0.9373 | 0.7007 | 1.245 |
hu_birth_cohort1775M1780 | 0.8213 | 0.6211 | 1.095 |
hu_birth_cohort1780M1785 | 0.7556 | 0.5717 | 1.001 |
hu_birth_cohort1785M1790 | 0.654 | 0.4982 | 0.8703 |
hu_birth_cohort1790M1795 | 0.7273 | 0.5616 | 0.9449 |
hu_birth_cohort1795M1800 | 0.6281 | 0.487 | 0.8102 |
hu_birth_cohort1800M1805 | 0.5814 | 0.4493 | 0.7461 |
hu_birth_cohort1805M1810 | 0.7611 | 0.5924 | 0.9772 |
hu_birth_cohort1810M1815 | 0.6438 | 0.5113 | 0.8127 |
hu_birth_cohort1815M1820 | 0.4937 | 0.3919 | 0.6171 |
hu_birth_cohort1820M1825 | 0.5908 | 0.4664 | 0.7423 |
hu_birth_cohort1825M1830 | 0.5785 | 0.4592 | 0.7226 |
hu_birth_cohort1830M1835 | 0.5729 | 0.4547 | 0.7258 |
hu_male1 | 1.317 | 1.203 | 1.438 |
hu_maternalage.factor1420 | 1.277 | 0.8085 | 1.995 |
hu_maternalage.factor3550 | 1.146 | 0.989 | 1.327 |
hu_paternalage_at_1st_sib | 1.078 | 0.9308 | 1.255 |
hu_paternalage.mean | 0.7968 | 0.5913 | 1.077 |
hu_paternal_loss01 | 1.771 | 1.225 | 2.576 |
hu_paternal_loss15 | 1.696 | 1.316 | 2.195 |
hu_paternal_loss510 | 1.216 | 0.9768 | 1.519 |
hu_paternal_loss1015 | 1.171 | 0.9406 | 1.449 |
hu_paternal_loss1520 | 1.107 | 0.9004 | 1.339 |
hu_paternal_loss2025 | 1.164 | 0.9596 | 1.416 |
hu_paternal_loss2530 | 1.062 | 0.8863 | 1.272 |
hu_paternal_loss3035 | 0.9752 | 0.8103 | 1.166 |
hu_paternal_loss3540 | 0.984 | 0.821 | 1.175 |
hu_paternal_loss4045 | 1.152 | 0.9484 | 1.386 |
hu_maternal_loss01 | 4.823 | 3.377 | 6.93 |
hu_maternal_loss15 | 1.79 | 1.415 | 2.26 |
hu_maternal_loss510 | 1.614 | 1.309 | 1.991 |
hu_maternal_loss1015 | 1.601 | 1.276 | 1.98 |
hu_maternal_loss1520 | 1.367 | 1.107 | 1.682 |
hu_maternal_loss2025 | 1.299 | 1.061 | 1.58 |
hu_maternal_loss2530 | 1.209 | 1.003 | 1.443 |
hu_maternal_loss3035 | 1.243 | 1.049 | 1.476 |
hu_maternal_loss3540 | 1.073 | 0.9145 | 1.261 |
hu_maternal_loss4045 | 1.322 | 1.109 | 1.556 |
hu_older_siblings1 | 0.9893 | 0.854 | 1.151 |
hu_older_siblings2 | 0.8571 | 0.7097 | 1.04 |
hu_older_siblings3 | 0.824 | 0.6519 | 1.05 |
hu_older_siblings4 | 0.7953 | 0.5906 | 1.076 |
hu_older_siblings5P | 0.5647 | 0.3828 | 0.8455 |
hu_nr.siblings | 1.114 | 1.074 | 1.156 |
hu_last_born1 | 1.084 | 0.9581 | 1.226 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.28 | [1.92;2.66] | [2.05;2.52] |
estimate father 35y | 2.09 | [1.67;2.62] | [1.81;2.41] |
percentage change | -8.15 | [-24.89;12.04] | [-19.48;4.45] |
OR/IRR | 1.07 | [0.96;1.19] | [1;1.15] |
OR hurdle | 1.3 | [0.97;1.7] | [1.07;1.56] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r13_control_paternal_afb.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
Most of the previous literature has not used multilevel modelling, but linear group fixed effects (essentially dummy variables on the many thousands of families in the model). We believe our multilevel modelling approach has the advantage of allowing us to examine the effect of including predictors at the level of the family in the same model.
This allows us to
a) appropriately model a zero-inflated outcome such as number of children including those who died young (we’re not aware of a linear group fixed effect approach that handles hurdle or zero-inflated models)
b) examine group-level slopes for paternal age and potentially to examine moderators at the level of the family (though we did not do this)
c) explicitly model confounders at the level of the family (e.g. number of siblings).
Nevertheless, the prevalence of this approach in the literature mandates that we show how our approach compares. We fit this model using the R package “lfe” and the function felm. All covariates that were not estimable in principle were removed (i.e. number of siblings, paternalage.mean).
##
## Call:
## felm(formula = children ~ paternalage + birth_cohort + male + maternalage.factor + paternal_loss + maternal_loss + older_siblings + last_born | idParents, data = model_data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -6.540 -1.573 -0.406 1.027 13.206
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## paternalage 0.17133 0.33109 0.52 0.6049
## birth_cohort1760-1765 0.35779 0.30698 1.17 0.2438
## birth_cohort1765-1770 0.43853 0.36777 1.19 0.2331
## birth_cohort1770-1775 0.07183 0.43056 0.17 0.8675
## birth_cohort1775-1780 0.33986 0.49995 0.68 0.4967
## birth_cohort1780-1785 0.29355 0.57618 0.51 0.6104
## birth_cohort1785-1790 0.53056 0.65420 0.81 0.4174
## birth_cohort1790-1795 0.24865 0.72729 0.34 0.7325
## birth_cohort1795-1800 0.40321 0.79871 0.50 0.6137
## birth_cohort1800-1805 0.43146 0.87013 0.50 0.6200
## birth_cohort1805-1810 0.04673 0.94056 0.05 0.9604
## birth_cohort1810-1815 0.28026 1.01517 0.28 0.7825
## birth_cohort1815-1820 0.45538 1.08829 0.42 0.6756
## birth_cohort1820-1825 0.10329 1.16289 0.09 0.9292
## birth_cohort1825-1830 0.02640 1.23637 0.02 0.9830
## birth_cohort1830-1835 0.08661 1.31567 0.07 0.9475
## male1 -0.14997 0.06556 -2.29 0.0222 *
## maternalage.factor(14,20] -0.00410 0.34424 -0.01 0.9905
## maternalage.factor(35,50] -0.07796 0.12193 -0.64 0.5226
## paternal_loss[0,1] -1.21851 0.86270 -1.41 0.1579
## paternal_loss(1,5] -0.87654 0.78516 -1.12 0.2643
## paternal_loss(5,10] -0.57513 0.70735 -0.81 0.4162
## paternal_loss(10,15] -0.33064 0.62346 -0.53 0.5959
## paternal_loss(15,20] -0.40807 0.54130 -0.75 0.4510
## paternal_loss(20,25] -0.46128 0.46058 -1.00 0.3166
## paternal_loss(25,30] -0.14990 0.38148 -0.39 0.6944
## paternal_loss(30,35] 0.05866 0.30667 0.19 0.8483
## paternal_loss(35,40] -0.00628 0.23313 -0.03 0.9785
## paternal_loss(40,45] -0.16950 0.18583 -0.91 0.3617
## maternal_loss[0,1] -2.61540 0.76190 -3.43 0.0006 ***
## maternal_loss(1,5] -1.44548 0.69856 -2.07 0.0386 *
## maternal_loss(5,10] -1.15962 0.63265 -1.83 0.0668 .
## maternal_loss(10,15] -1.06668 0.56336 -1.89 0.0583 .
## maternal_loss(15,20] -1.04251 0.49101 -2.12 0.0338 *
## maternal_loss(20,25] -0.80468 0.41611 -1.93 0.0532 .
## maternal_loss(25,30] -0.90035 0.34067 -2.64 0.0082 **
## maternal_loss(30,35] -0.82516 0.27191 -3.03 0.0024 **
## maternal_loss(35,40] -0.46819 0.20562 -2.28 0.0228 *
## maternal_loss(40,45] -0.59883 0.16080 -3.72 0.0002 ***
## older_siblings1 0.09345 0.10355 0.90 0.3669
## older_siblings2 0.10412 0.13844 0.75 0.4520
## older_siblings3 0.06926 0.17858 0.39 0.6982
## older_siblings4 0.07289 0.22201 0.33 0.7427
## older_siblings5+ 0.43075 0.29032 1.48 0.1379
## last_born1 -0.14407 0.08445 -1.71 0.0880 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.78 on 7216 degrees of freedom
## Multiple R-squared(full model): 0.279 Adjusted R-squared: 0.0557
## Multiple R-squared(proj model): 0.0144 Adjusted R-squared: -0.29
## F-statistic(full model):1.25 on 2230 and 7216 DF, p-value: 1.56e-11
## F-statistic(proj model): 2.34 on 45 and 7216 DF, p-value: 1.16e-06
In this model we attempted allow for regional variation in paternal age effects and attempted to better control residual variation. Our approach was two-fold: to moderate paternal age by region and to add a random effect for the church parish in which the individual was born. However, for the modern Swedish data, we had no geographic data and no regional information, so this model was not fit.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents) + (1 | gebortk)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents) + (1 | gebortk)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~gebortk (Number of levels: 105)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.03 0.02 0.00 0.06 496 1.01
## sd(hu_Intercept) 0.13 0.05 0.04 0.23 978 1.00
##
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.22 0.01 0.20 0.25 1127 1.00
## sd(hu_Intercept) 0.46 0.05 0.37 0.54 691 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.65 0.08 1.50 1.81 1142
## paternalage 0.07 0.05 -0.04 0.17 930
## birth_cohort1760M1765 0.00 0.06 -0.13 0.12 1212
## birth_cohort1765M1770 -0.12 0.06 -0.23 -0.01 976
## birth_cohort1770M1775 -0.11 0.06 -0.22 0.00 996
## birth_cohort1775M1780 -0.03 0.06 -0.13 0.08 907
## birth_cohort1780M1785 -0.11 0.06 -0.22 0.00 877
## birth_cohort1785M1790 -0.10 0.06 -0.21 0.01 895
## birth_cohort1790M1795 -0.08 0.05 -0.18 0.03 838
## birth_cohort1795M1800 -0.11 0.05 -0.20 -0.01 655
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.02 642
## birth_cohort1805M1810 -0.14 0.05 -0.24 -0.04 580
## birth_cohort1810M1815 -0.10 0.05 -0.19 0.00 636
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 562
## birth_cohort1820M1825 -0.19 0.05 -0.28 -0.10 563
## birth_cohort1825M1830 -0.21 0.05 -0.30 -0.12 443
## birth_cohort1830M1835 -0.18 0.05 -0.27 -0.09 571
## male1 0.08 0.02 0.04 0.11 3000
## maternalage.factor1420 -0.05 0.09 -0.24 0.13 3000
## maternalage.factor3550 0.00 0.03 -0.06 0.05 3000
## paternalage.mean -0.07 0.06 -0.18 0.04 938
## paternal_loss01 -0.15 0.07 -0.30 -0.01 3000
## paternal_loss15 -0.04 0.05 -0.14 0.06 1974
## paternal_loss510 -0.07 0.04 -0.15 0.02 1229
## paternal_loss1015 0.01 0.04 -0.07 0.08 1419
## paternal_loss1520 -0.09 0.04 -0.17 -0.02 1406
## paternal_loss2025 -0.12 0.04 -0.19 -0.05 1400
## paternal_loss2530 -0.01 0.04 -0.08 0.06 1283
## paternal_loss3035 -0.03 0.03 -0.10 0.04 1330
## paternal_loss3540 -0.01 0.03 -0.07 0.05 1545
## paternal_loss4045 -0.01 0.04 -0.08 0.06 3000
## maternal_loss01 0.10 0.08 -0.06 0.25 3000
## maternal_loss15 -0.02 0.05 -0.12 0.07 3000
## maternal_loss510 0.07 0.04 -0.01 0.15 3000
## maternal_loss1015 0.03 0.04 -0.05 0.11 3000
## maternal_loss1520 0.00 0.04 -0.08 0.08 3000
## maternal_loss2025 0.00 0.04 -0.07 0.08 3000
## maternal_loss2530 -0.02 0.03 -0.09 0.04 3000
## maternal_loss3035 -0.05 0.03 -0.11 0.01 1848
## maternal_loss3540 -0.03 0.03 -0.09 0.02 2283
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## older_siblings1 0.03 0.03 -0.03 0.08 1669
## older_siblings2 -0.05 0.04 -0.12 0.02 971
## older_siblings3 -0.08 0.05 -0.17 0.02 916
## older_siblings4 -0.09 0.06 -0.21 0.02 844
## older_siblings5P -0.10 0.08 -0.25 0.05 865
## nr.siblings 0.01 0.01 0.00 0.02 1107
## last_born1 -0.04 0.02 -0.09 0.00 3000
## hu_Intercept -0.35 0.20 -0.73 0.05 1309
## hu_paternalage 0.27 0.14 -0.01 0.54 923
## hu_birth_cohort1760M1765 -0.03 0.17 -0.35 0.29 3000
## hu_birth_cohort1765M1770 -0.29 0.14 -0.57 -0.01 1058
## hu_birth_cohort1770M1775 -0.05 0.15 -0.33 0.23 1038
## hu_birth_cohort1775M1780 -0.19 0.14 -0.47 0.09 1085
## hu_birth_cohort1780M1785 -0.26 0.15 -0.55 0.02 1105
## hu_birth_cohort1785M1790 -0.41 0.14 -0.69 -0.13 943
## hu_birth_cohort1790M1795 -0.30 0.14 -0.57 -0.05 924
## hu_birth_cohort1795M1800 -0.45 0.13 -0.71 -0.21 882
## hu_birth_cohort1800M1805 -0.52 0.13 -0.76 -0.27 783
## hu_birth_cohort1805M1810 -0.25 0.13 -0.51 -0.01 782
## hu_birth_cohort1810M1815 -0.42 0.12 -0.67 -0.19 797
## hu_birth_cohort1815M1820 -0.69 0.12 -0.92 -0.46 691
## hu_birth_cohort1820M1825 -0.50 0.12 -0.73 -0.27 714
## hu_birth_cohort1825M1830 -0.52 0.12 -0.76 -0.29 740
## hu_birth_cohort1830M1835 -0.53 0.12 -0.78 -0.30 762
## hu_male1 0.27 0.04 0.19 0.36 3000
## hu_maternalage.factor1420 0.24 0.24 -0.21 0.70 3000
## hu_maternalage.factor3550 0.13 0.07 -0.01 0.27 3000
## hu_paternalage.mean -0.18 0.15 -0.47 0.10 969
## hu_paternal_loss01 0.56 0.19 0.19 0.93 3000
## hu_paternal_loss15 0.51 0.13 0.25 0.77 3000
## hu_paternal_loss510 0.17 0.11 -0.04 0.39 1537
## hu_paternal_loss1015 0.14 0.10 -0.07 0.34 1422
## hu_paternal_loss1520 0.09 0.10 -0.11 0.29 1297
## hu_paternal_loss2025 0.14 0.10 -0.04 0.34 1256
## hu_paternal_loss2530 0.05 0.09 -0.14 0.23 1346
## hu_paternal_loss3035 -0.04 0.09 -0.21 0.15 1322
## hu_paternal_loss3540 -0.03 0.09 -0.20 0.15 1514
## hu_paternal_loss4045 0.14 0.10 -0.06 0.32 3000
## hu_maternal_loss01 1.57 0.19 1.20 1.93 3000
## hu_maternal_loss15 0.58 0.12 0.35 0.81 3000
## hu_maternal_loss510 0.48 0.11 0.27 0.69 3000
## hu_maternal_loss1015 0.47 0.11 0.25 0.68 3000
## hu_maternal_loss1520 0.31 0.11 0.10 0.51 3000
## hu_maternal_loss2025 0.26 0.11 0.05 0.46 3000
## hu_maternal_loss2530 0.19 0.09 0.01 0.36 3000
## hu_maternal_loss3035 0.21 0.09 0.04 0.38 3000
## hu_maternal_loss3540 0.07 0.08 -0.09 0.22 3000
## hu_maternal_loss4045 0.27 0.09 0.09 0.44 3000
## hu_older_siblings1 -0.01 0.07 -0.16 0.13 1719
## hu_older_siblings2 -0.15 0.10 -0.34 0.03 1137
## hu_older_siblings3 -0.19 0.13 -0.44 0.06 1011
## hu_older_siblings4 -0.23 0.16 -0.53 0.08 1001
## hu_older_siblings5P -0.57 0.21 -0.97 -0.16 927
## hu_nr.siblings 0.10 0.02 0.07 0.14 1253
## hu_last_born1 0.08 0.06 -0.04 0.20 3000
## Rhat
## Intercept 1.00
## paternalage 1.01
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.00
## birth_cohort1770M1775 1.00
## birth_cohort1775M1780 1.00
## birth_cohort1780M1785 1.00
## birth_cohort1785M1790 1.00
## birth_cohort1790M1795 1.01
## birth_cohort1795M1800 1.00
## birth_cohort1800M1805 1.01
## birth_cohort1805M1810 1.01
## birth_cohort1810M1815 1.00
## birth_cohort1815M1820 1.01
## birth_cohort1820M1825 1.01
## birth_cohort1825M1830 1.01
## birth_cohort1830M1835 1.00
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.01
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.01
## older_siblings3 1.01
## older_siblings4 1.01
## older_siblings5P 1.01
## nr.siblings 1.01
## last_born1 1.00
## hu_Intercept 1.00
## hu_paternalage 1.01
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.00
## hu_birth_cohort1775M1780 1.00
## hu_birth_cohort1780M1785 1.00
## hu_birth_cohort1785M1790 1.00
## hu_birth_cohort1790M1795 1.00
## hu_birth_cohort1795M1800 1.00
## hu_birth_cohort1800M1805 1.00
## hu_birth_cohort1805M1810 1.00
## hu_birth_cohort1810M1815 1.00
## hu_birth_cohort1815M1820 1.00
## hu_birth_cohort1820M1825 1.00
## hu_birth_cohort1825M1830 1.00
## hu_birth_cohort1830M1835 1.00
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.01
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_older_siblings1 1.01
## hu_older_siblings2 1.01
## hu_older_siblings3 1.01
## hu_older_siblings4 1.01
## hu_older_siblings5P 1.01
## hu_nr.siblings 1.01
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.227 | 4.461 | 6.138 |
paternalage | 1.072 | 0.9626 | 1.189 |
birth_cohort1760M1765 | 0.9958 | 0.8817 | 1.126 |
birth_cohort1765M1770 | 0.8879 | 0.7924 | 0.9931 |
birth_cohort1770M1775 | 0.8921 | 0.8015 | 0.9972 |
birth_cohort1775M1780 | 0.9744 | 0.874 | 1.081 |
birth_cohort1780M1785 | 0.8965 | 0.8029 | 1.002 |
birth_cohort1785M1790 | 0.9088 | 0.8131 | 1.009 |
birth_cohort1790M1795 | 0.9277 | 0.8357 | 1.027 |
birth_cohort1795M1800 | 0.9003 | 0.8187 | 0.9938 |
birth_cohort1800M1805 | 0.8928 | 0.8126 | 0.9821 |
birth_cohort1805M1810 | 0.8717 | 0.7897 | 0.9593 |
birth_cohort1810M1815 | 0.9061 | 0.8258 | 0.9959 |
birth_cohort1815M1820 | 0.8665 | 0.7917 | 0.9471 |
birth_cohort1820M1825 | 0.8298 | 0.7595 | 0.9071 |
birth_cohort1825M1830 | 0.811 | 0.741 | 0.888 |
birth_cohort1830M1835 | 0.8334 | 0.7609 | 0.9144 |
male1 | 1.081 | 1.046 | 1.117 |
maternalage.factor1420 | 0.9499 | 0.7892 | 1.139 |
maternalage.factor3550 | 0.9969 | 0.9445 | 1.055 |
paternalage.mean | 0.931 | 0.8345 | 1.037 |
paternal_loss01 | 0.8576 | 0.7428 | 0.995 |
paternal_loss15 | 0.9615 | 0.8675 | 1.065 |
paternal_loss510 | 0.9337 | 0.8572 | 1.019 |
paternal_loss1015 | 1.005 | 0.9282 | 1.087 |
paternal_loss1520 | 0.9099 | 0.8434 | 0.9817 |
paternal_loss2025 | 0.8865 | 0.8233 | 0.956 |
paternal_loss2530 | 0.9904 | 0.9224 | 1.063 |
paternal_loss3035 | 0.9715 | 0.9083 | 1.036 |
paternal_loss3540 | 0.9914 | 0.93 | 1.055 |
paternal_loss4045 | 0.9908 | 0.9234 | 1.063 |
maternal_loss01 | 1.101 | 0.9446 | 1.283 |
maternal_loss15 | 0.9786 | 0.8882 | 1.07 |
maternal_loss510 | 1.073 | 0.9907 | 1.163 |
maternal_loss1015 | 1.026 | 0.9488 | 1.113 |
maternal_loss1520 | 1.003 | 0.925 | 1.086 |
maternal_loss2025 | 1.003 | 0.9321 | 1.079 |
maternal_loss2530 | 0.9774 | 0.9121 | 1.044 |
maternal_loss3035 | 0.9487 | 0.8918 | 1.013 |
maternal_loss3540 | 0.9659 | 0.9147 | 1.023 |
maternal_loss4045 | 0.9706 | 0.912 | 1.031 |
older_siblings1 | 1.025 | 0.9711 | 1.083 |
older_siblings2 | 0.9525 | 0.8895 | 1.023 |
older_siblings3 | 0.9274 | 0.846 | 1.017 |
older_siblings4 | 0.9102 | 0.8102 | 1.02 |
older_siblings5P | 0.9056 | 0.781 | 1.051 |
nr.siblings | 1.01 | 0.996 | 1.025 |
last_born1 | 0.9569 | 0.9143 | 1.002 |
hu_Intercept | 0.7074 | 0.4821 | 1.048 |
hu_paternalage | 1.304 | 0.9868 | 1.724 |
hu_birth_cohort1760M1765 | 0.9683 | 0.7052 | 1.338 |
hu_birth_cohort1765M1770 | 0.7449 | 0.5639 | 0.9879 |
hu_birth_cohort1770M1775 | 0.9527 | 0.7196 | 1.262 |
hu_birth_cohort1775M1780 | 0.8269 | 0.6264 | 1.096 |
hu_birth_cohort1780M1785 | 0.7684 | 0.5764 | 1.018 |
hu_birth_cohort1785M1790 | 0.663 | 0.4994 | 0.8808 |
hu_birth_cohort1790M1795 | 0.7382 | 0.5656 | 0.9557 |
hu_birth_cohort1795M1800 | 0.6351 | 0.4904 | 0.8069 |
hu_birth_cohort1800M1805 | 0.5961 | 0.4674 | 0.767 |
hu_birth_cohort1805M1810 | 0.7764 | 0.5993 | 0.9945 |
hu_birth_cohort1810M1815 | 0.658 | 0.5135 | 0.8289 |
hu_birth_cohort1815M1820 | 0.5035 | 0.3969 | 0.6309 |
hu_birth_cohort1820M1825 | 0.6051 | 0.48 | 0.7604 |
hu_birth_cohort1825M1830 | 0.592 | 0.4675 | 0.7448 |
hu_birth_cohort1830M1835 | 0.5868 | 0.4571 | 0.7382 |
hu_male1 | 1.315 | 1.208 | 1.434 |
hu_maternalage.factor1420 | 1.268 | 0.8066 | 2.015 |
hu_maternalage.factor3550 | 1.139 | 0.9936 | 1.308 |
hu_paternalage.mean | 0.8352 | 0.6269 | 1.101 |
hu_paternal_loss01 | 1.742 | 1.215 | 2.527 |
hu_paternal_loss15 | 1.662 | 1.29 | 2.169 |
hu_paternal_loss510 | 1.189 | 0.9597 | 1.482 |
hu_paternal_loss1015 | 1.151 | 0.9365 | 1.41 |
hu_paternal_loss1520 | 1.09 | 0.8982 | 1.336 |
hu_paternal_loss2025 | 1.152 | 0.9618 | 1.404 |
hu_paternal_loss2530 | 1.053 | 0.8699 | 1.262 |
hu_paternal_loss3035 | 0.9656 | 0.8092 | 1.16 |
hu_paternal_loss3540 | 0.9749 | 0.8151 | 1.162 |
hu_paternal_loss4045 | 1.148 | 0.9387 | 1.384 |
hu_maternal_loss01 | 4.797 | 3.323 | 6.887 |
hu_maternal_loss15 | 1.788 | 1.414 | 2.256 |
hu_maternal_loss510 | 1.62 | 1.315 | 1.985 |
hu_maternal_loss1015 | 1.594 | 1.286 | 1.974 |
hu_maternal_loss1520 | 1.359 | 1.102 | 1.673 |
hu_maternal_loss2025 | 1.291 | 1.051 | 1.582 |
hu_maternal_loss2530 | 1.206 | 1.008 | 1.437 |
hu_maternal_loss3035 | 1.235 | 1.042 | 1.459 |
hu_maternal_loss3540 | 1.068 | 0.9123 | 1.248 |
hu_maternal_loss4045 | 1.316 | 1.1 | 1.557 |
hu_older_siblings1 | 0.9891 | 0.855 | 1.139 |
hu_older_siblings2 | 0.8565 | 0.7112 | 1.034 |
hu_older_siblings3 | 0.8231 | 0.6469 | 1.065 |
hu_older_siblings4 | 0.7959 | 0.5868 | 1.082 |
hu_older_siblings5P | 0.5641 | 0.3803 | 0.8498 |
hu_nr.siblings | 1.107 | 1.068 | 1.149 |
hu_last_born1 | 1.082 | 0.9575 | 1.219 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.31 | [1.94;2.72] | [2.07;2.58] |
estimate father 35y | 2.11 | [1.68;2.65] | [1.81;2.45] |
percentage change | -8.31 | [-24.61;10.99] | [-19.6;4.1] |
OR/IRR | 1.07 | [0.96;1.19] | [1;1.15] |
OR hurdle | 1.3 | [0.99;1.72] | [1.08;1.56] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r15_region_moderator_parish_ranef.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
Only in the DDB (historical Swedish data), parishes in some of the regions were still unlinked. This means that individuals could occur in more than one parish and not be linked. However, the region of Skellefteå was fully linked. Here, we test what happens when we restrict our dataset to Skellefteå.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9286)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2179)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.22 0.01 0.20 0.25 1145 1.01
## sd(hu_Intercept) 0.48 0.05 0.38 0.56 673 1.00
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.65 0.08 1.50 1.81 1440
## paternalage 0.07 0.05 -0.03 0.18 511
## birth_cohort1760M1765 0.00 0.06 -0.12 0.12 1433
## birth_cohort1765M1770 -0.12 0.06 -0.24 0.00 992
## birth_cohort1770M1775 -0.11 0.06 -0.22 0.00 1003
## birth_cohort1775M1780 -0.03 0.05 -0.13 0.08 1037
## birth_cohort1780M1785 -0.11 0.06 -0.22 0.00 862
## birth_cohort1785M1790 -0.10 0.05 -0.20 0.01 913
## birth_cohort1790M1795 -0.08 0.05 -0.18 0.03 616
## birth_cohort1795M1800 -0.11 0.05 -0.20 -0.01 775
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.02 854
## birth_cohort1805M1810 -0.13 0.05 -0.23 -0.04 816
## birth_cohort1810M1815 -0.10 0.05 -0.19 -0.01 665
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 485
## birth_cohort1820M1825 -0.19 0.05 -0.28 -0.10 442
## birth_cohort1825M1830 -0.21 0.05 -0.30 -0.12 643
## birth_cohort1830M1835 -0.18 0.05 -0.28 -0.09 618
## male1 0.08 0.02 0.05 0.11 3000
## maternalage.factor1420 -0.05 0.09 -0.23 0.13 3000
## maternalage.factor3550 0.00 0.03 -0.06 0.05 3000
## paternalage.mean -0.07 0.06 -0.18 0.04 543
## paternal_loss01 -0.15 0.07 -0.30 0.00 3000
## paternal_loss15 -0.04 0.05 -0.13 0.06 1903
## paternal_loss510 -0.07 0.04 -0.15 0.01 1674
## paternal_loss1015 0.01 0.04 -0.07 0.09 1598
## paternal_loss1520 -0.09 0.04 -0.17 -0.02 1588
## paternal_loss2025 -0.12 0.04 -0.19 -0.05 1604
## paternal_loss2530 -0.01 0.03 -0.07 0.06 1419
## paternal_loss3035 -0.03 0.03 -0.09 0.03 1537
## paternal_loss3540 -0.01 0.03 -0.07 0.05 1718
## paternal_loss4045 -0.01 0.04 -0.08 0.06 3000
## maternal_loss01 0.10 0.08 -0.06 0.26 3000
## maternal_loss15 -0.02 0.05 -0.11 0.07 1870
## maternal_loss510 0.07 0.04 -0.01 0.15 1801
## maternal_loss1015 0.03 0.04 -0.06 0.11 1974
## maternal_loss1520 0.00 0.04 -0.08 0.08 3000
## maternal_loss2025 0.00 0.04 -0.07 0.08 3000
## maternal_loss2530 -0.02 0.04 -0.09 0.05 1925
## maternal_loss3035 -0.05 0.03 -0.12 0.01 2035
## maternal_loss3540 -0.03 0.03 -0.09 0.02 2104
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## older_siblings1 0.03 0.03 -0.03 0.08 1316
## older_siblings2 -0.05 0.04 -0.12 0.02 788
## older_siblings3 -0.08 0.05 -0.17 0.02 618
## older_siblings4 -0.09 0.06 -0.22 0.02 593
## older_siblings5P -0.10 0.08 -0.26 0.05 565
## nr.siblings 0.01 0.01 0.00 0.02 636
## last_born1 -0.04 0.02 -0.09 0.00 3000
## hu_Intercept -0.33 0.19 -0.71 0.04 1158
## hu_paternalage 0.07 0.15 -0.23 0.36 848
## hu_birth_cohort1760M1765 -0.07 0.16 -0.39 0.24 3000
## hu_birth_cohort1765M1770 -0.30 0.15 -0.58 -0.03 1203
## hu_birth_cohort1770M1775 -0.05 0.14 -0.33 0.23 1022
## hu_birth_cohort1775M1780 -0.20 0.14 -0.48 0.07 1058
## hu_birth_cohort1780M1785 -0.28 0.14 -0.56 0.00 1146
## hu_birth_cohort1785M1790 -0.44 0.14 -0.69 -0.16 1043
## hu_birth_cohort1790M1795 -0.30 0.13 -0.56 -0.03 943
## hu_birth_cohort1795M1800 -0.46 0.12 -0.71 -0.21 876
## hu_birth_cohort1800M1805 -0.52 0.12 -0.77 -0.28 925
## hu_birth_cohort1805M1810 -0.28 0.12 -0.51 -0.04 912
## hu_birth_cohort1810M1815 -0.45 0.12 -0.69 -0.20 846
## hu_birth_cohort1815M1820 -0.70 0.11 -0.92 -0.47 747
## hu_birth_cohort1820M1825 -0.52 0.11 -0.74 -0.29 740
## hu_birth_cohort1825M1830 -0.54 0.12 -0.77 -0.32 789
## hu_birth_cohort1830M1835 -0.55 0.12 -0.77 -0.32 820
## hu_male1 0.27 0.05 0.18 0.36 3000
## hu_maternalage.factor1420 0.23 0.23 -0.21 0.71 3000
## hu_maternalage.factor3550 0.08 0.07 -0.06 0.23 3000
## hu_paternalage.mean 0.00 0.15 -0.30 0.29 853
## hu_paternal_loss01 0.63 0.18 0.29 0.99 3000
## hu_paternal_loss15 0.56 0.13 0.31 0.82 1703
## hu_paternal_loss510 0.21 0.11 -0.01 0.43 1617
## hu_paternal_loss1015 0.17 0.11 -0.04 0.37 1456
## hu_paternal_loss1520 0.11 0.10 -0.10 0.31 1500
## hu_paternal_loss2025 0.16 0.10 -0.03 0.36 1384
## hu_paternal_loss2530 0.07 0.09 -0.11 0.25 1330
## hu_paternal_loss3035 -0.01 0.09 -0.18 0.17 1356
## hu_paternal_loss3540 -0.01 0.09 -0.19 0.16 1293
## hu_paternal_loss4045 0.15 0.10 -0.05 0.35 3000
## hu_maternal_loss01 1.61 0.19 1.25 1.98 3000
## hu_maternal_loss15 0.58 0.12 0.35 0.83 3000
## hu_maternal_loss510 0.47 0.11 0.26 0.68 3000
## hu_maternal_loss1015 0.46 0.11 0.25 0.67 3000
## hu_maternal_loss1520 0.32 0.11 0.10 0.52 3000
## hu_maternal_loss2025 0.27 0.10 0.07 0.47 3000
## hu_maternal_loss2530 0.19 0.09 0.01 0.37 3000
## hu_maternal_loss3035 0.20 0.09 0.03 0.37 3000
## hu_maternal_loss3540 0.07 0.08 -0.09 0.22 3000
## hu_maternal_loss4045 0.28 0.09 0.11 0.46 3000
## hu_older_siblings1 0.03 0.08 -0.11 0.19 1696
## hu_older_siblings2 -0.06 0.10 -0.26 0.13 979
## hu_older_siblings3 -0.07 0.13 -0.31 0.18 870
## hu_older_siblings4 -0.06 0.16 -0.38 0.26 907
## hu_older_siblings5P -0.40 0.22 -0.82 0.02 823
## hu_nr.siblings 0.09 0.02 0.05 0.13 996
## hu_last_born1 0.12 0.06 -0.01 0.25 3000
## Rhat
## Intercept 1.00
## paternalage 1.01
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.01
## birth_cohort1770M1775 1.01
## birth_cohort1775M1780 1.01
## birth_cohort1780M1785 1.01
## birth_cohort1785M1790 1.01
## birth_cohort1790M1795 1.01
## birth_cohort1795M1800 1.01
## birth_cohort1800M1805 1.01
## birth_cohort1805M1810 1.01
## birth_cohort1810M1815 1.01
## birth_cohort1815M1820 1.01
## birth_cohort1820M1825 1.01
## birth_cohort1825M1830 1.01
## birth_cohort1830M1835 1.01
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.01
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.01
## older_siblings3 1.01
## older_siblings4 1.01
## older_siblings5P 1.01
## nr.siblings 1.01
## last_born1 1.00
## hu_Intercept 1.01
## hu_paternalage 1.00
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.00
## hu_birth_cohort1775M1780 1.00
## hu_birth_cohort1780M1785 1.00
## hu_birth_cohort1785M1790 1.00
## hu_birth_cohort1790M1795 1.00
## hu_birth_cohort1795M1800 1.00
## hu_birth_cohort1800M1805 1.00
## hu_birth_cohort1805M1810 1.00
## hu_birth_cohort1810M1815 1.00
## hu_birth_cohort1815M1820 1.01
## hu_birth_cohort1820M1825 1.01
## hu_birth_cohort1825M1830 1.00
## hu_birth_cohort1830M1835 1.01
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.00
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.00
## hu_older_siblings3 1.00
## hu_older_siblings4 1.00
## hu_older_siblings5P 1.00
## hu_nr.siblings 1.00
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.231 | 4.503 | 6.124 |
paternalage | 1.071 | 0.9663 | 1.195 |
birth_cohort1760M1765 | 0.9974 | 0.8831 | 1.129 |
birth_cohort1765M1770 | 0.8876 | 0.7895 | 0.9978 |
birth_cohort1770M1775 | 0.8915 | 0.7988 | 0.9954 |
birth_cohort1775M1780 | 0.9746 | 0.8763 | 1.083 |
birth_cohort1780M1785 | 0.8971 | 0.8022 | 0.9998 |
birth_cohort1785M1790 | 0.9083 | 0.8174 | 1.011 |
birth_cohort1790M1795 | 0.9276 | 0.8365 | 1.029 |
birth_cohort1795M1800 | 0.8999 | 0.8198 | 0.9926 |
birth_cohort1800M1805 | 0.8937 | 0.8121 | 0.981 |
birth_cohort1805M1810 | 0.8742 | 0.795 | 0.962 |
birth_cohort1810M1815 | 0.9073 | 0.827 | 0.9928 |
birth_cohort1815M1820 | 0.8679 | 0.7939 | 0.9497 |
birth_cohort1820M1825 | 0.831 | 0.7592 | 0.9062 |
birth_cohort1825M1830 | 0.8111 | 0.7413 | 0.8872 |
birth_cohort1830M1835 | 0.8337 | 0.759 | 0.9117 |
male1 | 1.081 | 1.046 | 1.116 |
maternalage.factor1420 | 0.9489 | 0.7947 | 1.134 |
maternalage.factor3550 | 0.9961 | 0.9426 | 1.051 |
paternalage.mean | 0.9311 | 0.8346 | 1.038 |
paternal_loss01 | 0.8601 | 0.7426 | 0.9979 |
paternal_loss15 | 0.965 | 0.8741 | 1.066 |
paternal_loss510 | 0.9345 | 0.8612 | 1.015 |
paternal_loss1015 | 1.008 | 0.9322 | 1.089 |
paternal_loss1520 | 0.9113 | 0.8447 | 0.9797 |
paternal_loss2025 | 0.8877 | 0.8263 | 0.9511 |
paternal_loss2530 | 0.9912 | 0.9295 | 1.06 |
paternal_loss3035 | 0.9728 | 0.9122 | 1.035 |
paternal_loss3540 | 0.9914 | 0.9339 | 1.053 |
paternal_loss4045 | 0.991 | 0.9253 | 1.065 |
maternal_loss01 | 1.104 | 0.9406 | 1.298 |
maternal_loss15 | 0.9797 | 0.8926 | 1.074 |
maternal_loss510 | 1.072 | 0.9929 | 1.158 |
maternal_loss1015 | 1.027 | 0.946 | 1.112 |
maternal_loss1520 | 1.002 | 0.9244 | 1.084 |
maternal_loss2025 | 1.004 | 0.9303 | 1.079 |
maternal_loss2530 | 0.9782 | 0.9124 | 1.052 |
maternal_loss3035 | 0.9491 | 0.8899 | 1.01 |
maternal_loss3540 | 0.9657 | 0.9117 | 1.019 |
maternal_loss4045 | 0.9701 | 0.9118 | 1.03 |
older_siblings1 | 1.026 | 0.9736 | 1.082 |
older_siblings2 | 0.9524 | 0.885 | 1.023 |
older_siblings3 | 0.9276 | 0.8431 | 1.016 |
older_siblings4 | 0.9099 | 0.8057 | 1.019 |
older_siblings5P | 0.9065 | 0.7679 | 1.055 |
nr.siblings | 1.01 | 0.9956 | 1.025 |
last_born1 | 0.9576 | 0.9134 | 1.001 |
hu_Intercept | 0.7159 | 0.491 | 1.045 |
hu_paternalage | 1.07 | 0.7962 | 1.429 |
hu_birth_cohort1760M1765 | 0.9338 | 0.6767 | 1.275 |
hu_birth_cohort1765M1770 | 0.7397 | 0.5599 | 0.9723 |
hu_birth_cohort1770M1775 | 0.9469 | 0.7163 | 1.258 |
hu_birth_cohort1775M1780 | 0.8172 | 0.6186 | 1.073 |
hu_birth_cohort1780M1785 | 0.7563 | 0.5702 | 0.9979 |
hu_birth_cohort1785M1790 | 0.647 | 0.5003 | 0.8539 |
hu_birth_cohort1790M1795 | 0.7442 | 0.5735 | 0.97 |
hu_birth_cohort1795M1800 | 0.6291 | 0.4941 | 0.8103 |
hu_birth_cohort1800M1805 | 0.5935 | 0.4637 | 0.7553 |
hu_birth_cohort1805M1810 | 0.7568 | 0.5984 | 0.9617 |
hu_birth_cohort1810M1815 | 0.638 | 0.503 | 0.8148 |
hu_birth_cohort1815M1820 | 0.4982 | 0.4 | 0.6248 |
hu_birth_cohort1820M1825 | 0.5966 | 0.4766 | 0.7474 |
hu_birth_cohort1825M1830 | 0.5799 | 0.4609 | 0.7274 |
hu_birth_cohort1830M1835 | 0.5787 | 0.462 | 0.7241 |
hu_male1 | 1.316 | 1.197 | 1.437 |
hu_maternalage.factor1420 | 1.256 | 0.8138 | 2.043 |
hu_maternalage.factor3550 | 1.087 | 0.9417 | 1.261 |
hu_paternalage.mean | 0.9972 | 0.7414 | 1.343 |
hu_paternal_loss01 | 1.877 | 1.331 | 2.693 |
hu_paternal_loss15 | 1.756 | 1.358 | 2.268 |
hu_paternal_loss510 | 1.233 | 0.9872 | 1.543 |
hu_paternal_loss1015 | 1.181 | 0.9561 | 1.451 |
hu_paternal_loss1520 | 1.118 | 0.9042 | 1.364 |
hu_paternal_loss2025 | 1.179 | 0.9697 | 1.438 |
hu_paternal_loss2530 | 1.07 | 0.8954 | 1.282 |
hu_paternal_loss3035 | 0.9927 | 0.8347 | 1.19 |
hu_paternal_loss3540 | 0.9904 | 0.8282 | 1.174 |
hu_paternal_loss4045 | 1.165 | 0.95 | 1.413 |
hu_maternal_loss01 | 4.98 | 3.504 | 7.254 |
hu_maternal_loss15 | 1.795 | 1.421 | 2.287 |
hu_maternal_loss510 | 1.605 | 1.299 | 1.978 |
hu_maternal_loss1015 | 1.58 | 1.288 | 1.956 |
hu_maternal_loss1520 | 1.379 | 1.107 | 1.689 |
hu_maternal_loss2025 | 1.309 | 1.072 | 1.594 |
hu_maternal_loss2530 | 1.206 | 1.007 | 1.451 |
hu_maternal_loss3035 | 1.22 | 1.03 | 1.441 |
hu_maternal_loss3540 | 1.067 | 0.9155 | 1.246 |
hu_maternal_loss4045 | 1.329 | 1.116 | 1.582 |
hu_older_siblings1 | 1.034 | 0.8926 | 1.204 |
hu_older_siblings2 | 0.9402 | 0.7725 | 1.144 |
hu_older_siblings3 | 0.9356 | 0.7325 | 1.196 |
hu_older_siblings4 | 0.9411 | 0.6816 | 1.3 |
hu_older_siblings5P | 0.6683 | 0.4392 | 1.025 |
hu_nr.siblings | 1.096 | 1.055 | 1.138 |
hu_last_born1 | 1.126 | 0.9935 | 1.281 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.21 | [1.87;2.58] | [1.98;2.45] |
estimate father 35y | 2.27 | [1.8;2.8] | [1.96;2.61] |
percentage change | 2.67 | [-15.43;25.56] | [-10.09;17.18] |
OR/IRR | 1.07 | [0.97;1.2] | [1;1.15] |
OR hurdle | 1.07 | [0.8;1.43] | [0.88;1.29] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r17_simulate_downs.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
To make models computationally feasible and because early mortality was negligible, we fit the very large modern Swedish dataset with a poisson()
family distribution. All historical datasets had high early mortality, so we thought a hurdle_poisson()
was more appropriate. Here, we show what happens when we reverse this. The hurdle_poisson()
model can be fit to the modern Swedish data here, because we only use a subset.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.9 0.02 0.86 0.94 638 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 0.80 0.15 0.50 1.08 626
## paternalage -0.02 0.06 -0.14 0.10 595
## birth_cohort1760M1765 0.16 0.07 0.01 0.30 820
## birth_cohort1765M1770 0.23 0.07 0.09 0.37 534
## birth_cohort1770M1775 0.03 0.08 -0.12 0.17 429
## birth_cohort1775M1780 0.21 0.08 0.06 0.35 356
## birth_cohort1780M1785 0.20 0.08 0.04 0.35 373
## birth_cohort1785M1790 0.32 0.08 0.17 0.47 336
## birth_cohort1790M1795 0.22 0.08 0.07 0.37 314
## birth_cohort1795M1800 0.31 0.07 0.16 0.46 297
## birth_cohort1800M1805 0.34 0.07 0.20 0.48 319
## birth_cohort1805M1810 0.15 0.07 0.02 0.30 301
## birth_cohort1810M1815 0.29 0.07 0.15 0.43 273
## birth_cohort1815M1820 0.40 0.07 0.26 0.54 253
## birth_cohort1820M1825 0.25 0.08 0.10 0.39 254
## birth_cohort1825M1830 0.22 0.08 0.07 0.37 243
## birth_cohort1830M1835 0.26 0.08 0.10 0.42 257
## male1 -0.08 0.02 -0.12 -0.05 3000
## maternalage.factor1420 -0.04 0.10 -0.23 0.15 3000
## maternalage.factor3550 -0.07 0.03 -0.13 -0.01 3000
## paternalage.mean -0.05 0.07 -0.18 0.08 621
## paternal_loss01 -0.61 0.10 -0.81 -0.42 575
## paternal_loss15 -0.39 0.08 -0.55 -0.24 409
## paternal_loss510 -0.22 0.07 -0.36 -0.08 410
## paternal_loss1015 -0.09 0.06 -0.22 0.03 388
## paternal_loss1520 -0.15 0.06 -0.27 -0.04 414
## paternal_loss2025 -0.19 0.05 -0.30 -0.08 490
## paternal_loss2530 -0.03 0.05 -0.13 0.07 516
## paternal_loss3035 0.05 0.04 -0.04 0.14 580
## paternal_loss3540 0.02 0.04 -0.05 0.10 776
## paternal_loss4045 -0.07 0.04 -0.15 0.01 1186
## maternal_loss01 -1.32 0.10 -1.52 -1.12 736
## maternal_loss15 -0.50 0.07 -0.65 -0.36 447
## maternal_loss510 -0.34 0.06 -0.47 -0.21 418
## maternal_loss1015 -0.33 0.06 -0.46 -0.21 430
## maternal_loss1520 -0.33 0.06 -0.45 -0.21 501
## maternal_loss2025 -0.25 0.06 -0.36 -0.14 519
## maternal_loss2530 -0.29 0.05 -0.39 -0.19 516
## maternal_loss3035 -0.29 0.04 -0.38 -0.21 590
## maternal_loss3540 -0.15 0.04 -0.23 -0.08 679
## maternal_loss4045 -0.26 0.03 -0.32 -0.19 1510
## older_siblings1 0.06 0.03 0.00 0.11 1174
## older_siblings2 0.07 0.04 0.00 0.14 701
## older_siblings3 0.06 0.05 -0.03 0.16 663
## older_siblings4 0.06 0.06 -0.07 0.18 648
## older_siblings5P 0.25 0.08 0.10 0.41 661
## nr.siblings -0.02 0.01 -0.04 0.00 452
## last_born1 -0.08 0.02 -0.13 -0.04 3000
## Rhat
## Intercept 1.01
## paternalage 1.00
## birth_cohort1760M1765 1.01
## birth_cohort1765M1770 1.01
## birth_cohort1770M1775 1.01
## birth_cohort1775M1780 1.01
## birth_cohort1780M1785 1.01
## birth_cohort1785M1790 1.02
## birth_cohort1790M1795 1.01
## birth_cohort1795M1800 1.02
## birth_cohort1800M1805 1.01
## birth_cohort1805M1810 1.02
## birth_cohort1810M1815 1.02
## birth_cohort1815M1820 1.03
## birth_cohort1820M1825 1.03
## birth_cohort1825M1830 1.04
## birth_cohort1830M1835 1.04
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.01
## paternal_loss01 1.01
## paternal_loss15 1.02
## paternal_loss510 1.02
## paternal_loss1015 1.02
## paternal_loss1520 1.02
## paternal_loss2025 1.01
## paternal_loss2530 1.01
## paternal_loss3035 1.01
## paternal_loss3540 1.01
## paternal_loss4045 1.01
## maternal_loss01 1.00
## maternal_loss15 1.01
## maternal_loss510 1.01
## maternal_loss1015 1.01
## maternal_loss1520 1.01
## maternal_loss2025 1.01
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.00
## older_siblings3 1.00
## older_siblings4 1.01
## older_siblings5P 1.01
## nr.siblings 1.01
## last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 2.217 | 1.651 | 2.938 |
paternalage | 0.9788 | 0.8696 | 1.104 |
birth_cohort1760M1765 | 1.169 | 1.012 | 1.345 |
birth_cohort1765M1770 | 1.253 | 1.091 | 1.445 |
birth_cohort1770M1775 | 1.028 | 0.8913 | 1.19 |
birth_cohort1775M1780 | 1.228 | 1.057 | 1.425 |
birth_cohort1780M1785 | 1.22 | 1.045 | 1.416 |
birth_cohort1785M1790 | 1.38 | 1.181 | 1.597 |
birth_cohort1790M1795 | 1.249 | 1.072 | 1.447 |
birth_cohort1795M1800 | 1.366 | 1.179 | 1.576 |
birth_cohort1800M1805 | 1.406 | 1.219 | 1.62 |
birth_cohort1805M1810 | 1.167 | 1.016 | 1.346 |
birth_cohort1810M1815 | 1.339 | 1.166 | 1.542 |
birth_cohort1815M1820 | 1.493 | 1.296 | 1.715 |
birth_cohort1820M1825 | 1.278 | 1.105 | 1.479 |
birth_cohort1825M1830 | 1.249 | 1.077 | 1.446 |
birth_cohort1830M1835 | 1.295 | 1.103 | 1.517 |
male1 | 0.9187 | 0.8886 | 0.9488 |
maternalage.factor1420 | 0.9654 | 0.7981 | 1.159 |
maternalage.factor3550 | 0.9316 | 0.8773 | 0.9891 |
paternalage.mean | 0.9506 | 0.8313 | 1.086 |
paternal_loss01 | 0.5437 | 0.4455 | 0.6582 |
paternal_loss15 | 0.6771 | 0.5787 | 0.7883 |
paternal_loss510 | 0.8056 | 0.6977 | 0.92 |
paternal_loss1015 | 0.9121 | 0.8005 | 1.032 |
paternal_loss1520 | 0.8574 | 0.7601 | 0.9622 |
paternal_loss2025 | 0.8238 | 0.7399 | 0.9187 |
paternal_loss2530 | 0.9704 | 0.8788 | 1.074 |
paternal_loss3035 | 1.051 | 0.9621 | 1.148 |
paternal_loss3540 | 1.025 | 0.9487 | 1.106 |
paternal_loss4045 | 0.9355 | 0.8647 | 1.012 |
maternal_loss01 | 0.2665 | 0.2193 | 0.326 |
maternal_loss15 | 0.604 | 0.5244 | 0.7007 |
maternal_loss510 | 0.7104 | 0.6254 | 0.8082 |
maternal_loss1015 | 0.7157 | 0.6326 | 0.8091 |
maternal_loss1520 | 0.7186 | 0.6389 | 0.8097 |
maternal_loss2025 | 0.7819 | 0.6989 | 0.8716 |
maternal_loss2530 | 0.7479 | 0.6793 | 0.8243 |
maternal_loss3035 | 0.7464 | 0.6849 | 0.812 |
maternal_loss3540 | 0.8577 | 0.7984 | 0.9232 |
maternal_loss4045 | 0.7741 | 0.7237 | 0.8288 |
older_siblings1 | 1.057 | 1.003 | 1.113 |
older_siblings2 | 1.074 | 0.9985 | 1.153 |
older_siblings3 | 1.066 | 0.9737 | 1.172 |
older_siblings4 | 1.059 | 0.9351 | 1.194 |
older_siblings5P | 1.29 | 1.106 | 1.508 |
nr.siblings | 0.9763 | 0.9561 | 0.9974 |
last_born1 | 0.9201 | 0.8801 | 0.9613 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 1.57 | [1.35;1.81] | [1.42;1.72] |
estimate father 35y | 1.53 | [1.27;1.84] | [1.36;1.73] |
percentage change | -2.09 | [-13.04;10.4] | [-9.43;5.83] |
OR/IRR | 0.98 | [0.87;1.1] | [0.91;1.06] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r18_hurdle_poisson.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
Previous analysts sometimes decided to use the normal distribution to predict (potentially zero-inflated) count data. Here, we refit our models using a normal distribution for the outcome. We show that estimates for the paternal age effect can be estimated to have a substantially different magnitude, because of this, but did not change direction.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: gaussian(identity)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## sigma ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.53 0.06 0.4 0.65 304 1.03
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 2.92 0.25 2.44 3.39 1574
## paternalage -0.18 0.18 -0.52 0.18 987
## birth_cohort1760M1765 0.08 0.21 -0.33 0.47 1492
## birth_cohort1765M1770 0.12 0.19 -0.25 0.49 1241
## birth_cohort1770M1775 -0.10 0.18 -0.45 0.26 1336
## birth_cohort1775M1780 0.20 0.18 -0.16 0.56 1212
## birth_cohort1780M1785 0.10 0.19 -0.28 0.47 1220
## birth_cohort1785M1790 0.30 0.18 -0.07 0.65 1107
## birth_cohort1790M1795 0.23 0.17 -0.10 0.56 1138
## birth_cohort1795M1800 0.32 0.17 -0.01 0.64 1099
## birth_cohort1800M1805 0.39 0.16 0.07 0.71 1007
## birth_cohort1805M1810 0.08 0.16 -0.24 0.40 950
## birth_cohort1810M1815 0.33 0.16 0.02 0.63 934
## birth_cohort1815M1820 0.51 0.15 0.23 0.80 956
## birth_cohort1820M1825 0.25 0.15 -0.04 0.55 838
## birth_cohort1825M1830 0.23 0.15 -0.06 0.53 849
## birth_cohort1830M1835 0.28 0.16 -0.04 0.59 926
## male1 -0.15 0.06 -0.26 -0.03 3000
## maternalage.factor1420 -0.35 0.30 -0.94 0.23 3000
## maternalage.factor3550 -0.12 0.09 -0.30 0.07 3000
## paternalage.mean 0.09 0.18 -0.28 0.44 983
## paternal_loss01 -0.82 0.22 -1.25 -0.39 3000
## paternal_loss15 -0.58 0.16 -0.89 -0.26 3000
## paternal_loss510 -0.35 0.15 -0.63 -0.06 1720
## paternal_loss1015 -0.20 0.14 -0.48 0.07 1626
## paternal_loss1520 -0.31 0.13 -0.57 -0.06 932
## paternal_loss2025 -0.40 0.13 -0.65 -0.16 1606
## paternal_loss2530 -0.10 0.12 -0.34 0.14 1130
## paternal_loss3035 -0.07 0.12 -0.31 0.17 1445
## paternal_loss3540 -0.05 0.12 -0.28 0.18 1658
## paternal_loss4045 -0.20 0.13 -0.46 0.06 3000
## maternal_loss01 -1.37 0.20 -1.77 -0.99 3000
## maternal_loss15 -0.61 0.15 -0.90 -0.31 3000
## maternal_loss510 -0.41 0.14 -0.68 -0.13 3000
## maternal_loss1015 -0.43 0.13 -0.70 -0.18 3000
## maternal_loss1520 -0.33 0.14 -0.59 -0.06 3000
## maternal_loss2025 -0.25 0.14 -0.51 0.02 3000
## maternal_loss2530 -0.24 0.12 -0.49 -0.01 3000
## maternal_loss3035 -0.32 0.12 -0.54 -0.09 3000
## maternal_loss3540 -0.13 0.11 -0.34 0.08 2139
## maternal_loss4045 -0.35 0.11 -0.57 -0.12 3000
## older_siblings1 0.06 0.10 -0.14 0.26 3000
## older_siblings2 0.10 0.12 -0.14 0.34 1254
## older_siblings3 0.09 0.16 -0.22 0.39 1022
## older_siblings4 0.08 0.19 -0.29 0.47 1056
## older_siblings5P 0.45 0.26 -0.07 0.94 887
## nr.siblings -0.09 0.02 -0.14 -0.05 1201
## last_born1 -0.17 0.08 -0.33 -0.01 3000
## Rhat
## Intercept 1.00
## paternalage 1.01
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.01
## birth_cohort1770M1775 1.01
## birth_cohort1775M1780 1.01
## birth_cohort1780M1785 1.01
## birth_cohort1785M1790 1.01
## birth_cohort1790M1795 1.01
## birth_cohort1795M1800 1.01
## birth_cohort1800M1805 1.01
## birth_cohort1805M1810 1.01
## birth_cohort1810M1815 1.01
## birth_cohort1815M1820 1.01
## birth_cohort1820M1825 1.01
## birth_cohort1825M1830 1.01
## birth_cohort1830M1835 1.01
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.01
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.00
## older_siblings3 1.01
## older_siblings4 1.00
## older_siblings5P 1.01
## nr.siblings 1.00
## last_born1 1.00
##
## Family Specific Parameters:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sigma 2.79 0.02 2.75 2.84 3000 1.01
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 18.59 | 11.43 | 29.62 |
paternalage | 0.8388 | 0.5955 | 1.199 |
birth_cohort1760M1765 | 1.08 | 0.7173 | 1.607 |
birth_cohort1765M1770 | 1.129 | 0.7775 | 1.63 |
birth_cohort1770M1775 | 0.9087 | 0.6347 | 1.292 |
birth_cohort1775M1780 | 1.225 | 0.8549 | 1.744 |
birth_cohort1780M1785 | 1.107 | 0.7594 | 1.601 |
birth_cohort1785M1790 | 1.352 | 0.9288 | 1.922 |
birth_cohort1790M1795 | 1.26 | 0.9078 | 1.752 |
birth_cohort1795M1800 | 1.37 | 0.991 | 1.89 |
birth_cohort1800M1805 | 1.482 | 1.075 | 2.035 |
birth_cohort1805M1810 | 1.088 | 0.7857 | 1.485 |
birth_cohort1810M1815 | 1.386 | 1.022 | 1.869 |
birth_cohort1815M1820 | 1.673 | 1.253 | 2.232 |
birth_cohort1820M1825 | 1.28 | 0.9561 | 1.738 |
birth_cohort1825M1830 | 1.263 | 0.941 | 1.691 |
birth_cohort1830M1835 | 1.328 | 0.9653 | 1.796 |
male1 | 0.8633 | 0.7691 | 0.966 |
maternalage.factor1420 | 0.7049 | 0.3897 | 1.257 |
maternalage.factor3550 | 0.8897 | 0.7432 | 1.072 |
paternalage.mean | 1.089 | 0.7587 | 1.556 |
paternal_loss01 | 0.4417 | 0.286 | 0.6778 |
paternal_loss15 | 0.5614 | 0.4096 | 0.7693 |
paternal_loss510 | 0.702 | 0.5301 | 0.9383 |
paternal_loss1015 | 0.8223 | 0.6162 | 1.076 |
paternal_loss1520 | 0.734 | 0.5653 | 0.9422 |
paternal_loss2025 | 0.67 | 0.5223 | 0.8551 |
paternal_loss2530 | 0.9053 | 0.7135 | 1.15 |
paternal_loss3035 | 0.9356 | 0.7354 | 1.182 |
paternal_loss3540 | 0.9546 | 0.7529 | 1.202 |
paternal_loss4045 | 0.8226 | 0.6305 | 1.064 |
maternal_loss01 | 0.2539 | 0.1703 | 0.3723 |
maternal_loss15 | 0.5435 | 0.405 | 0.7333 |
maternal_loss510 | 0.6664 | 0.5047 | 0.877 |
maternal_loss1015 | 0.65 | 0.4986 | 0.837 |
maternal_loss1520 | 0.7176 | 0.5522 | 0.9414 |
maternal_loss2025 | 0.7773 | 0.5985 | 1.017 |
maternal_loss2530 | 0.7846 | 0.6154 | 0.9939 |
maternal_loss3035 | 0.7277 | 0.5818 | 0.9156 |
maternal_loss3540 | 0.874 | 0.7145 | 1.088 |
maternal_loss4045 | 0.7076 | 0.5638 | 0.8857 |
older_siblings1 | 1.064 | 0.8691 | 1.292 |
older_siblings2 | 1.104 | 0.8705 | 1.401 |
older_siblings3 | 1.091 | 0.8003 | 1.473 |
older_siblings4 | 1.087 | 0.7465 | 1.599 |
older_siblings5P | 1.562 | 0.9287 | 2.566 |
nr.siblings | 0.912 | 0.8715 | 0.9543 |
last_born1 | 0.8457 | 0.7169 | 0.9935 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.31 | [2;2.63] | [2.11;2.52] |
estimate father 35y | 2.13 | [1.76;2.53] | [1.89;2.39] |
percentage change | -7.69 | [-21.51;8.34] | [-17.23;2.43] |
OR/IRR | 0.84 | [0.6;1.2] | [0.66;1.05] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r19_normal_distribution.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
In this model, we test what happens when we do not adjust for maternal age, because it is highly collinear with paternal age.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + birth_cohort + male + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 1000; warmup = 500; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.22 0.01 0.20 0.25 1295 1.00
## sd(hu_Intercept) 0.47 0.05 0.38 0.56 562 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.65 0.08 1.50 1.81 1198
## paternalage 0.07 0.05 -0.03 0.16 859
## birth_cohort1760M1765 0.00 0.06 -0.13 0.12 1368
## birth_cohort1765M1770 -0.12 0.06 -0.23 -0.01 981
## birth_cohort1770M1775 -0.11 0.06 -0.23 0.00 889
## birth_cohort1775M1780 -0.03 0.06 -0.13 0.08 774
## birth_cohort1780M1785 -0.11 0.06 -0.22 0.00 903
## birth_cohort1785M1790 -0.10 0.06 -0.20 0.01 830
## birth_cohort1790M1795 -0.08 0.05 -0.18 0.03 783
## birth_cohort1795M1800 -0.11 0.05 -0.20 -0.01 755
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.02 630
## birth_cohort1805M1810 -0.14 0.05 -0.23 -0.04 665
## birth_cohort1810M1815 -0.10 0.05 -0.19 0.00 648
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 635
## birth_cohort1820M1825 -0.18 0.05 -0.27 -0.09 636
## birth_cohort1825M1830 -0.21 0.05 -0.30 -0.12 613
## birth_cohort1830M1835 -0.18 0.05 -0.27 -0.09 661
## male1 0.08 0.02 0.05 0.11 3000
## paternalage.mean -0.07 0.05 -0.17 0.03 915
## paternal_loss01 -0.15 0.07 -0.30 -0.01 3000
## paternal_loss15 -0.04 0.05 -0.14 0.06 2118
## paternal_loss510 -0.07 0.04 -0.16 0.01 1480
## paternal_loss1015 0.01 0.04 -0.07 0.08 1293
## paternal_loss1520 -0.09 0.04 -0.17 -0.02 1290
## paternal_loss2025 -0.12 0.04 -0.19 -0.05 1458
## paternal_loss2530 -0.01 0.03 -0.08 0.06 1174
## paternal_loss3035 -0.03 0.03 -0.09 0.04 1156
## paternal_loss3540 -0.01 0.03 -0.07 0.05 1268
## paternal_loss4045 -0.01 0.04 -0.09 0.06 1960
## maternal_loss01 0.10 0.08 -0.06 0.25 3000
## maternal_loss15 -0.02 0.05 -0.11 0.07 3000
## maternal_loss510 0.07 0.04 -0.01 0.15 3000
## maternal_loss1015 0.03 0.04 -0.06 0.10 3000
## maternal_loss1520 0.00 0.04 -0.08 0.08 3000
## maternal_loss2025 0.00 0.04 -0.07 0.08 3000
## maternal_loss2530 -0.02 0.03 -0.09 0.05 3000
## maternal_loss3035 -0.05 0.03 -0.12 0.01 3000
## maternal_loss3540 -0.03 0.03 -0.09 0.02 3000
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## older_siblings1 0.03 0.03 -0.02 0.08 1541
## older_siblings2 -0.05 0.03 -0.11 0.02 1001
## older_siblings3 -0.07 0.05 -0.16 0.02 957
## older_siblings4 -0.09 0.06 -0.20 0.02 914
## older_siblings5P -0.09 0.08 -0.25 0.06 830
## nr.siblings 0.01 0.01 0.00 0.02 1134
## last_born1 -0.04 0.02 -0.09 0.00 3000
## hu_Intercept -0.31 0.20 -0.69 0.07 1419
## hu_paternalage 0.36 0.14 0.10 0.62 860
## hu_birth_cohort1760M1765 -0.05 0.17 -0.38 0.28 3000
## hu_birth_cohort1765M1770 -0.32 0.15 -0.60 -0.03 1015
## hu_birth_cohort1770M1775 -0.07 0.14 -0.36 0.21 1055
## hu_birth_cohort1775M1780 -0.20 0.14 -0.47 0.07 1112
## hu_birth_cohort1780M1785 -0.27 0.14 -0.55 0.02 1390
## hu_birth_cohort1785M1790 -0.41 0.14 -0.68 -0.14 1430
## hu_birth_cohort1790M1795 -0.31 0.13 -0.56 -0.05 1360
## hu_birth_cohort1795M1800 -0.46 0.12 -0.71 -0.22 1133
## hu_birth_cohort1800M1805 -0.53 0.12 -0.76 -0.30 1208
## hu_birth_cohort1805M1810 -0.26 0.12 -0.50 -0.02 1109
## hu_birth_cohort1810M1815 -0.43 0.12 -0.66 -0.20 1003
## hu_birth_cohort1815M1820 -0.70 0.11 -0.92 -0.48 613
## hu_birth_cohort1820M1825 -0.53 0.11 -0.75 -0.30 1011
## hu_birth_cohort1825M1830 -0.54 0.11 -0.77 -0.32 712
## hu_birth_cohort1830M1835 -0.55 0.12 -0.78 -0.31 670
## hu_male1 0.27 0.05 0.19 0.37 3000
## hu_paternalage.mean -0.27 0.14 -0.54 0.00 846
## hu_paternal_loss01 0.55 0.18 0.21 0.91 3000
## hu_paternal_loss15 0.51 0.13 0.26 0.76 3000
## hu_paternal_loss510 0.17 0.11 -0.05 0.39 2029
## hu_paternal_loss1015 0.15 0.11 -0.06 0.35 1820
## hu_paternal_loss1520 0.09 0.10 -0.10 0.28 1654
## hu_paternal_loss2025 0.15 0.10 -0.04 0.34 1681
## hu_paternal_loss2530 0.06 0.09 -0.12 0.24 1736
## hu_paternal_loss3035 -0.03 0.09 -0.21 0.15 1662
## hu_paternal_loss3540 -0.02 0.09 -0.20 0.16 1668
## hu_paternal_loss4045 0.14 0.10 -0.06 0.34 3000
## hu_maternal_loss01 1.56 0.19 1.20 1.93 3000
## hu_maternal_loss15 0.57 0.12 0.34 0.80 3000
## hu_maternal_loss510 0.48 0.10 0.27 0.69 3000
## hu_maternal_loss1015 0.47 0.10 0.27 0.68 3000
## hu_maternal_loss1520 0.32 0.11 0.12 0.53 3000
## hu_maternal_loss2025 0.27 0.10 0.06 0.46 3000
## hu_maternal_loss2530 0.20 0.09 0.02 0.38 3000
## hu_maternal_loss3035 0.22 0.09 0.06 0.39 3000
## hu_maternal_loss3540 0.07 0.08 -0.09 0.24 3000
## hu_maternal_loss4045 0.28 0.08 0.12 0.44 3000
## hu_older_siblings1 -0.03 0.07 -0.18 0.11 3000
## hu_older_siblings2 -0.19 0.10 -0.38 0.00 1066
## hu_older_siblings3 -0.23 0.12 -0.47 0.01 962
## hu_older_siblings4 -0.26 0.16 -0.58 0.04 892
## hu_older_siblings5P -0.60 0.21 -1.02 -0.19 864
## hu_nr.siblings 0.11 0.02 0.07 0.14 1113
## hu_last_born1 0.09 0.06 -0.03 0.21 3000
## Rhat
## Intercept 1.00
## paternalage 1.00
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.01
## birth_cohort1770M1775 1.01
## birth_cohort1775M1780 1.01
## birth_cohort1780M1785 1.01
## birth_cohort1785M1790 1.01
## birth_cohort1790M1795 1.01
## birth_cohort1795M1800 1.01
## birth_cohort1800M1805 1.01
## birth_cohort1805M1810 1.01
## birth_cohort1810M1815 1.01
## birth_cohort1815M1820 1.01
## birth_cohort1820M1825 1.01
## birth_cohort1825M1830 1.01
## birth_cohort1830M1835 1.01
## male1 1.00
## paternalage.mean 1.00
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.00
## older_siblings3 1.00
## older_siblings4 1.00
## older_siblings5P 1.00
## nr.siblings 1.00
## last_born1 1.00
## hu_Intercept 1.00
## hu_paternalage 1.00
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.00
## hu_birth_cohort1775M1780 1.00
## hu_birth_cohort1780M1785 1.00
## hu_birth_cohort1785M1790 1.00
## hu_birth_cohort1790M1795 1.00
## hu_birth_cohort1795M1800 1.00
## hu_birth_cohort1800M1805 1.00
## hu_birth_cohort1805M1810 1.00
## hu_birth_cohort1810M1815 1.00
## hu_birth_cohort1815M1820 1.01
## hu_birth_cohort1820M1825 1.01
## hu_birth_cohort1825M1830 1.00
## hu_birth_cohort1830M1835 1.01
## hu_male1 1.00
## hu_paternalage.mean 1.00
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.00
## hu_older_siblings3 1.00
## hu_older_siblings4 1.00
## hu_older_siblings5P 1.00
## hu_nr.siblings 1.00
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.23 | 4.491 | 6.111 |
paternalage | 1.068 | 0.9739 | 1.179 |
birth_cohort1760M1765 | 0.9974 | 0.8797 | 1.132 |
birth_cohort1765M1770 | 0.8877 | 0.7946 | 0.9932 |
birth_cohort1770M1775 | 0.8916 | 0.7934 | 1.001 |
birth_cohort1775M1780 | 0.9742 | 0.8759 | 1.087 |
birth_cohort1780M1785 | 0.8955 | 0.7999 | 1.002 |
birth_cohort1785M1790 | 0.9076 | 0.8149 | 1.012 |
birth_cohort1790M1795 | 0.9272 | 0.838 | 1.031 |
birth_cohort1795M1800 | 0.8993 | 0.8147 | 0.9909 |
birth_cohort1800M1805 | 0.8928 | 0.8145 | 0.9786 |
birth_cohort1805M1810 | 0.8723 | 0.793 | 0.9618 |
birth_cohort1810M1815 | 0.9071 | 0.8307 | 0.9961 |
birth_cohort1815M1820 | 0.8674 | 0.7975 | 0.9497 |
birth_cohort1820M1825 | 0.8312 | 0.7612 | 0.9113 |
birth_cohort1825M1830 | 0.8109 | 0.7419 | 0.8868 |
birth_cohort1830M1835 | 0.8343 | 0.7631 | 0.9143 |
male1 | 1.081 | 1.046 | 1.117 |
paternalage.mean | 0.9338 | 0.8421 | 1.032 |
paternal_loss01 | 0.8578 | 0.7411 | 0.9874 |
paternal_loss15 | 0.9625 | 0.8685 | 1.061 |
paternal_loss510 | 0.9328 | 0.8562 | 1.014 |
paternal_loss1015 | 1.006 | 0.9294 | 1.089 |
paternal_loss1520 | 0.9096 | 0.8424 | 0.9831 |
paternal_loss2025 | 0.8861 | 0.8244 | 0.9529 |
paternal_loss2530 | 0.9901 | 0.9251 | 1.06 |
paternal_loss3035 | 0.9716 | 0.9102 | 1.04 |
paternal_loss3540 | 0.9902 | 0.9296 | 1.055 |
paternal_loss4045 | 0.9894 | 0.9181 | 1.063 |
maternal_loss01 | 1.105 | 0.9416 | 1.29 |
maternal_loss15 | 0.9801 | 0.8922 | 1.072 |
maternal_loss510 | 1.072 | 0.9919 | 1.162 |
maternal_loss1015 | 1.026 | 0.9459 | 1.108 |
maternal_loss1520 | 1.001 | 0.9211 | 1.087 |
maternal_loss2025 | 1.003 | 0.9308 | 1.082 |
maternal_loss2530 | 0.9772 | 0.9139 | 1.046 |
maternal_loss3035 | 0.95 | 0.8898 | 1.013 |
maternal_loss3540 | 0.966 | 0.913 | 1.023 |
maternal_loss4045 | 0.9707 | 0.9101 | 1.031 |
older_siblings1 | 1.028 | 0.9766 | 1.081 |
older_siblings2 | 0.9553 | 0.8941 | 1.022 |
older_siblings3 | 0.9303 | 0.8508 | 1.017 |
older_siblings4 | 0.9129 | 0.8158 | 1.022 |
older_siblings5P | 0.9095 | 0.7816 | 1.058 |
nr.siblings | 1.01 | 0.9959 | 1.024 |
last_born1 | 0.9572 | 0.916 | 1 |
hu_Intercept | 0.736 | 0.5014 | 1.076 |
hu_paternalage | 1.428 | 1.1 | 1.866 |
hu_birth_cohort1760M1765 | 0.9495 | 0.6866 | 1.327 |
hu_birth_cohort1765M1770 | 0.7294 | 0.5508 | 0.9695 |
hu_birth_cohort1770M1775 | 0.9345 | 0.6966 | 1.231 |
hu_birth_cohort1775M1780 | 0.819 | 0.6224 | 1.071 |
hu_birth_cohort1780M1785 | 0.7636 | 0.5748 | 1.025 |
hu_birth_cohort1785M1790 | 0.6631 | 0.5054 | 0.8691 |
hu_birth_cohort1790M1795 | 0.735 | 0.5693 | 0.9513 |
hu_birth_cohort1795M1800 | 0.6342 | 0.494 | 0.8038 |
hu_birth_cohort1800M1805 | 0.5878 | 0.4668 | 0.7387 |
hu_birth_cohort1805M1810 | 0.7686 | 0.6069 | 0.9775 |
hu_birth_cohort1810M1815 | 0.6489 | 0.5151 | 0.8171 |
hu_birth_cohort1815M1820 | 0.4973 | 0.3977 | 0.62 |
hu_birth_cohort1820M1825 | 0.5911 | 0.4703 | 0.7394 |
hu_birth_cohort1825M1830 | 0.5802 | 0.4651 | 0.7246 |
hu_birth_cohort1830M1835 | 0.5774 | 0.4574 | 0.73 |
hu_male1 | 1.315 | 1.205 | 1.442 |
hu_paternalage.mean | 0.7659 | 0.5808 | 1.005 |
hu_paternal_loss01 | 1.737 | 1.228 | 2.478 |
hu_paternal_loss15 | 1.668 | 1.292 | 2.137 |
hu_paternal_loss510 | 1.191 | 0.9544 | 1.474 |
hu_paternal_loss1015 | 1.16 | 0.9446 | 1.426 |
hu_paternal_loss1520 | 1.098 | 0.9019 | 1.329 |
hu_paternal_loss2025 | 1.161 | 0.9616 | 1.399 |
hu_paternal_loss2530 | 1.058 | 0.8845 | 1.273 |
hu_paternal_loss3035 | 0.9725 | 0.8101 | 1.16 |
hu_paternal_loss3540 | 0.9817 | 0.8179 | 1.174 |
hu_paternal_loss4045 | 1.147 | 0.9414 | 1.402 |
hu_maternal_loss01 | 4.776 | 3.316 | 6.906 |
hu_maternal_loss15 | 1.771 | 1.402 | 2.228 |
hu_maternal_loss510 | 1.616 | 1.311 | 1.985 |
hu_maternal_loss1015 | 1.606 | 1.316 | 1.969 |
hu_maternal_loss1520 | 1.375 | 1.124 | 1.697 |
hu_maternal_loss2025 | 1.305 | 1.064 | 1.584 |
hu_maternal_loss2530 | 1.222 | 1.02 | 1.466 |
hu_maternal_loss3035 | 1.251 | 1.063 | 1.479 |
hu_maternal_loss3540 | 1.076 | 0.9168 | 1.267 |
hu_maternal_loss4045 | 1.322 | 1.126 | 1.552 |
hu_older_siblings1 | 0.9677 | 0.8332 | 1.112 |
hu_older_siblings2 | 0.8249 | 0.6815 | 1.001 |
hu_older_siblings3 | 0.7932 | 0.6241 | 1.011 |
hu_older_siblings4 | 0.7675 | 0.5589 | 1.043 |
hu_older_siblings5P | 0.5469 | 0.3605 | 0.8273 |
hu_nr.siblings | 1.113 | 1.074 | 1.155 |
hu_last_born1 | 1.096 | 0.9714 | 1.236 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.32 | [1.96;2.69] | [2.08;2.56] |
estimate father 35y | 2 | [1.59;2.48] | [1.72;2.29] |
percentage change | -13.61 | [-29.41;3.73] | [-23.95;-2.63] |
OR/IRR | 1.07 | [0.97;1.18] | [1;1.14] |
OR hurdle | 1.43 | [1.1;1.87] | [1.2;1.7] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r20_no_maternalage_control.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
In this model, we adjust for maternal age using a continuous variable instead of three bins. This does not allow for nonlinear effects, but also does not aggregate the predictor. We cannot compare full siblings, test the effects of maternal and paternal age and adjust for average maternal and paternal age in the family (because the predictors are redundant), so that it is not perfectly possible to disentangle the contribution of maternal and paternal age and compare full siblings.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + maternalage + birth_cohort + male + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + maternalage + birth_cohort + male + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 1000; warmup = 500; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.23 0.01 0.20 0.25 976 1
## sd(hu_Intercept) 0.48 0.05 0.38 0.57 749 1
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.67 0.11 1.44 1.89 1119
## paternalage 0.07 0.06 -0.04 0.18 667
## maternalage -0.01 0.03 -0.06 0.05 1711
## birth_cohort1760M1765 0.00 0.06 -0.13 0.12 1096
## birth_cohort1765M1770 -0.12 0.06 -0.23 -0.01 780
## birth_cohort1770M1775 -0.11 0.06 -0.23 0.00 858
## birth_cohort1775M1780 -0.03 0.05 -0.13 0.08 735
## birth_cohort1780M1785 -0.11 0.06 -0.22 0.01 842
## birth_cohort1785M1790 -0.10 0.05 -0.20 0.01 737
## birth_cohort1790M1795 -0.07 0.05 -0.17 0.03 718
## birth_cohort1795M1800 -0.11 0.05 -0.20 -0.01 616
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.02 555
## birth_cohort1805M1810 -0.14 0.05 -0.23 -0.04 602
## birth_cohort1810M1815 -0.10 0.05 -0.19 -0.01 582
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 562
## birth_cohort1820M1825 -0.18 0.04 -0.27 -0.10 575
## birth_cohort1825M1830 -0.21 0.05 -0.30 -0.12 582
## birth_cohort1830M1835 -0.18 0.05 -0.27 -0.09 604
## male1 0.08 0.02 0.05 0.11 3000
## paternalage.mean -0.07 0.06 -0.19 0.03 723
## paternal_loss01 -0.16 0.08 -0.31 -0.02 3000
## paternal_loss15 -0.04 0.05 -0.15 0.06 1636
## paternal_loss510 -0.07 0.04 -0.15 0.01 1419
## paternal_loss1015 0.00 0.04 -0.07 0.08 1334
## paternal_loss1520 -0.10 0.04 -0.17 -0.02 1530
## paternal_loss2025 -0.12 0.04 -0.19 -0.05 1406
## paternal_loss2530 -0.01 0.03 -0.08 0.05 1376
## paternal_loss3035 -0.03 0.03 -0.10 0.04 1338
## paternal_loss3540 -0.01 0.03 -0.07 0.06 1468
## paternal_loss4045 -0.01 0.04 -0.08 0.06 2012
## maternal_loss01 0.10 0.08 -0.06 0.25 3000
## maternal_loss15 -0.02 0.05 -0.11 0.07 1800
## maternal_loss510 0.07 0.04 -0.01 0.15 1570
## maternal_loss1015 0.03 0.04 -0.06 0.11 1682
## maternal_loss1520 0.00 0.04 -0.08 0.08 1545
## maternal_loss2025 0.00 0.04 -0.07 0.08 1618
## maternal_loss2530 -0.02 0.04 -0.09 0.05 1565
## maternal_loss3035 -0.05 0.03 -0.12 0.02 1484
## maternal_loss3540 -0.03 0.03 -0.09 0.03 1588
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## older_siblings1 0.03 0.03 -0.03 0.08 1427
## older_siblings2 -0.05 0.04 -0.12 0.02 913
## older_siblings3 -0.07 0.05 -0.16 0.02 656
## older_siblings4 -0.09 0.06 -0.21 0.02 697
## older_siblings5P -0.09 0.08 -0.25 0.05 582
## nr.siblings 0.01 0.01 0.00 0.02 938
## last_born1 -0.04 0.02 -0.09 0.00 3000
## hu_Intercept -0.43 0.28 -0.97 0.14 1034
## hu_paternalage 0.31 0.15 0.02 0.63 705
## hu_maternalage 0.04 0.07 -0.10 0.18 1673
## hu_birth_cohort1760M1765 -0.06 0.17 -0.38 0.27 3000
## hu_birth_cohort1765M1770 -0.32 0.15 -0.61 -0.03 893
## hu_birth_cohort1770M1775 -0.07 0.14 -0.36 0.21 865
## hu_birth_cohort1775M1780 -0.20 0.14 -0.48 0.07 876
## hu_birth_cohort1780M1785 -0.28 0.15 -0.57 0.01 860
## hu_birth_cohort1785M1790 -0.42 0.14 -0.70 -0.13 843
## hu_birth_cohort1790M1795 -0.32 0.13 -0.58 -0.06 744
## hu_birth_cohort1795M1800 -0.46 0.13 -0.72 -0.21 702
## hu_birth_cohort1800M1805 -0.54 0.13 -0.79 -0.29 689
## hu_birth_cohort1805M1810 -0.27 0.13 -0.51 -0.01 676
## hu_birth_cohort1810M1815 -0.44 0.12 -0.68 -0.19 672
## hu_birth_cohort1815M1820 -0.70 0.12 -0.92 -0.47 599
## hu_birth_cohort1820M1825 -0.53 0.12 -0.76 -0.30 590
## hu_birth_cohort1825M1830 -0.55 0.12 -0.79 -0.32 597
## hu_birth_cohort1830M1835 -0.56 0.12 -0.80 -0.33 650
## hu_male1 0.27 0.05 0.18 0.36 3000
## hu_paternalage.mean -0.23 0.15 -0.54 0.06 708
## hu_paternal_loss01 0.57 0.18 0.21 0.94 3000
## hu_paternal_loss15 0.52 0.13 0.27 0.78 1348
## hu_paternal_loss510 0.19 0.11 -0.03 0.41 1107
## hu_paternal_loss1015 0.16 0.10 -0.05 0.37 1232
## hu_paternal_loss1520 0.10 0.10 -0.10 0.30 1142
## hu_paternal_loss2025 0.15 0.10 -0.04 0.35 1051
## hu_paternal_loss2530 0.06 0.09 -0.11 0.25 1167
## hu_paternal_loss3035 -0.02 0.09 -0.20 0.15 1391
## hu_paternal_loss3540 -0.01 0.09 -0.19 0.15 1517
## hu_paternal_loss4045 0.14 0.10 -0.05 0.34 1823
## hu_maternal_loss01 1.58 0.19 1.21 1.96 3000
## hu_maternal_loss15 0.58 0.12 0.35 0.82 3000
## hu_maternal_loss510 0.48 0.11 0.27 0.70 3000
## hu_maternal_loss1015 0.48 0.11 0.26 0.70 3000
## hu_maternal_loss1520 0.32 0.11 0.12 0.54 2153
## hu_maternal_loss2025 0.27 0.10 0.07 0.47 1977
## hu_maternal_loss2530 0.20 0.09 0.02 0.38 1866
## hu_maternal_loss3035 0.22 0.09 0.05 0.40 1774
## hu_maternal_loss3540 0.07 0.08 -0.09 0.23 1847
## hu_maternal_loss4045 0.28 0.09 0.11 0.44 1970
## hu_older_siblings1 -0.04 0.07 -0.18 0.11 1321
## hu_older_siblings2 -0.19 0.10 -0.38 -0.01 916
## hu_older_siblings3 -0.24 0.13 -0.48 0.01 788
## hu_older_siblings4 -0.27 0.16 -0.58 0.03 704
## hu_older_siblings5P -0.61 0.21 -1.02 -0.21 670
## hu_nr.siblings 0.11 0.02 0.07 0.14 845
## hu_last_born1 0.09 0.06 -0.03 0.21 3000
## Rhat
## Intercept 1.00
## paternalage 1.01
## maternalage 1.00
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.01
## birth_cohort1770M1775 1.01
## birth_cohort1775M1780 1.01
## birth_cohort1780M1785 1.01
## birth_cohort1785M1790 1.01
## birth_cohort1790M1795 1.01
## birth_cohort1795M1800 1.01
## birth_cohort1800M1805 1.01
## birth_cohort1805M1810 1.01
## birth_cohort1810M1815 1.01
## birth_cohort1815M1820 1.01
## birth_cohort1820M1825 1.01
## birth_cohort1825M1830 1.01
## birth_cohort1830M1835 1.01
## male1 1.00
## paternalage.mean 1.01
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.01
## older_siblings3 1.01
## older_siblings4 1.01
## older_siblings5P 1.01
## nr.siblings 1.01
## last_born1 1.00
## hu_Intercept 1.00
## hu_paternalage 1.01
## hu_maternalage 1.00
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.00
## hu_birth_cohort1775M1780 1.00
## hu_birth_cohort1780M1785 1.00
## hu_birth_cohort1785M1790 1.00
## hu_birth_cohort1790M1795 1.01
## hu_birth_cohort1795M1800 1.01
## hu_birth_cohort1800M1805 1.01
## hu_birth_cohort1805M1810 1.01
## hu_birth_cohort1810M1815 1.01
## hu_birth_cohort1815M1820 1.01
## hu_birth_cohort1820M1825 1.01
## hu_birth_cohort1825M1830 1.01
## hu_birth_cohort1830M1835 1.01
## hu_male1 1.00
## hu_paternalage.mean 1.01
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.01
## hu_paternal_loss1015 1.01
## hu_paternal_loss1520 1.01
## hu_paternal_loss2025 1.01
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.01
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.01
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.01
## hu_older_siblings3 1.01
## hu_older_siblings4 1.01
## hu_older_siblings5P 1.01
## hu_nr.siblings 1.01
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.311 | 4.219 | 6.591 |
paternalage | 1.074 | 0.9654 | 1.203 |
maternalage | 0.9941 | 0.9373 | 1.055 |
birth_cohort1760M1765 | 0.9977 | 0.8795 | 1.126 |
birth_cohort1765M1770 | 0.8889 | 0.7947 | 0.9942 |
birth_cohort1770M1775 | 0.8915 | 0.7963 | 0.9962 |
birth_cohort1775M1780 | 0.9748 | 0.8754 | 1.084 |
birth_cohort1780M1785 | 0.8983 | 0.8049 | 1.005 |
birth_cohort1785M1790 | 0.9093 | 0.8194 | 1.013 |
birth_cohort1790M1795 | 0.9292 | 0.8418 | 1.03 |
birth_cohort1795M1800 | 0.9002 | 0.8208 | 0.992 |
birth_cohort1800M1805 | 0.894 | 0.8136 | 0.9823 |
birth_cohort1805M1810 | 0.8732 | 0.793 | 0.9597 |
birth_cohort1810M1815 | 0.9076 | 0.8283 | 0.9947 |
birth_cohort1815M1820 | 0.8684 | 0.7953 | 0.9492 |
birth_cohort1820M1825 | 0.8313 | 0.7616 | 0.9059 |
birth_cohort1825M1830 | 0.8114 | 0.7432 | 0.8859 |
birth_cohort1830M1835 | 0.8336 | 0.7641 | 0.9169 |
male1 | 1.081 | 1.046 | 1.116 |
paternalage.mean | 0.9291 | 0.8276 | 1.034 |
paternal_loss01 | 0.8533 | 0.7335 | 0.9834 |
paternal_loss15 | 0.9606 | 0.8642 | 1.06 |
paternal_loss510 | 0.932 | 0.8569 | 1.015 |
paternal_loss1015 | 1.005 | 0.9304 | 1.086 |
paternal_loss1520 | 0.9087 | 0.8425 | 0.9798 |
paternal_loss2025 | 0.8855 | 0.8249 | 0.954 |
paternal_loss2530 | 0.9897 | 0.9223 | 1.055 |
paternal_loss3035 | 0.9708 | 0.9089 | 1.036 |
paternal_loss3540 | 0.9904 | 0.9298 | 1.057 |
paternal_loss4045 | 0.9896 | 0.9216 | 1.063 |
maternal_loss01 | 1.103 | 0.9402 | 1.287 |
maternal_loss15 | 0.9802 | 0.8951 | 1.076 |
maternal_loss510 | 1.071 | 0.9901 | 1.161 |
maternal_loss1015 | 1.027 | 0.9436 | 1.113 |
maternal_loss1520 | 1.002 | 0.9239 | 1.085 |
maternal_loss2025 | 1.004 | 0.9308 | 1.085 |
maternal_loss2530 | 0.9785 | 0.9145 | 1.046 |
maternal_loss3035 | 0.9503 | 0.8905 | 1.016 |
maternal_loss3540 | 0.9671 | 0.9097 | 1.026 |
maternal_loss4045 | 0.9712 | 0.9109 | 1.033 |
older_siblings1 | 1.028 | 0.9732 | 1.082 |
older_siblings2 | 0.9554 | 0.8911 | 1.024 |
older_siblings3 | 0.9314 | 0.8504 | 1.021 |
older_siblings4 | 0.913 | 0.8125 | 1.019 |
older_siblings5P | 0.911 | 0.7827 | 1.056 |
nr.siblings | 1.01 | 0.9957 | 1.024 |
last_born1 | 0.9566 | 0.9151 | 1 |
hu_Intercept | 0.6529 | 0.3794 | 1.147 |
hu_paternalage | 1.364 | 1.023 | 1.87 |
hu_maternalage | 1.046 | 0.9068 | 1.196 |
hu_birth_cohort1760M1765 | 0.9464 | 0.6864 | 1.308 |
hu_birth_cohort1765M1770 | 0.7254 | 0.5425 | 0.9724 |
hu_birth_cohort1770M1775 | 0.9284 | 0.6977 | 1.235 |
hu_birth_cohort1775M1780 | 0.8156 | 0.6178 | 1.074 |
hu_birth_cohort1780M1785 | 0.7562 | 0.5636 | 1.014 |
hu_birth_cohort1785M1790 | 0.6573 | 0.4952 | 0.8753 |
hu_birth_cohort1790M1795 | 0.7293 | 0.5598 | 0.9426 |
hu_birth_cohort1795M1800 | 0.629 | 0.4858 | 0.8067 |
hu_birth_cohort1800M1805 | 0.5816 | 0.4525 | 0.7463 |
hu_birth_cohort1805M1810 | 0.7645 | 0.6003 | 0.9866 |
hu_birth_cohort1810M1815 | 0.6467 | 0.5066 | 0.8239 |
hu_birth_cohort1815M1820 | 0.4954 | 0.3978 | 0.6237 |
hu_birth_cohort1820M1825 | 0.5895 | 0.4694 | 0.7419 |
hu_birth_cohort1825M1830 | 0.5757 | 0.4548 | 0.7294 |
hu_birth_cohort1830M1835 | 0.5726 | 0.4497 | 0.7219 |
hu_male1 | 1.314 | 1.199 | 1.436 |
hu_paternalage.mean | 0.7972 | 0.5818 | 1.063 |
hu_paternal_loss01 | 1.764 | 1.235 | 2.551 |
hu_paternal_loss15 | 1.689 | 1.31 | 2.182 |
hu_paternal_loss510 | 1.207 | 0.9706 | 1.508 |
hu_paternal_loss1015 | 1.169 | 0.9549 | 1.444 |
hu_paternal_loss1520 | 1.106 | 0.9074 | 1.347 |
hu_paternal_loss2025 | 1.167 | 0.9611 | 1.413 |
hu_paternal_loss2530 | 1.067 | 0.8936 | 1.281 |
hu_paternal_loss3035 | 0.9755 | 0.8161 | 1.165 |
hu_paternal_loss3540 | 0.9855 | 0.8297 | 1.167 |
hu_paternal_loss4045 | 1.152 | 0.9504 | 1.409 |
hu_maternal_loss01 | 4.85 | 3.365 | 7.131 |
hu_maternal_loss15 | 1.785 | 1.415 | 2.266 |
hu_maternal_loss510 | 1.623 | 1.304 | 2.015 |
hu_maternal_loss1015 | 1.612 | 1.303 | 2.009 |
hu_maternal_loss1520 | 1.372 | 1.123 | 1.713 |
hu_maternal_loss2025 | 1.304 | 1.069 | 1.595 |
hu_maternal_loss2530 | 1.219 | 1.017 | 1.467 |
hu_maternal_loss3035 | 1.248 | 1.055 | 1.486 |
hu_maternal_loss3540 | 1.073 | 0.9106 | 1.263 |
hu_maternal_loss4045 | 1.32 | 1.113 | 1.558 |
hu_older_siblings1 | 0.9656 | 0.8392 | 1.114 |
hu_older_siblings2 | 0.8242 | 0.6808 | 0.9935 |
hu_older_siblings3 | 0.7896 | 0.6167 | 1.005 |
hu_older_siblings4 | 0.7642 | 0.5622 | 1.03 |
hu_older_siblings5P | 0.5443 | 0.3603 | 0.8102 |
hu_nr.siblings | 1.113 | 1.074 | 1.154 |
hu_last_born1 | 1.096 | 0.9703 | 1.24 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.24 | [1.86;2.7] | [1.98;2.53] |
estimate father 35y | 1.99 | [1.6;2.48] | [1.73;2.31] |
percentage change | -10.86 | [-28.22;8.72] | [-22.22;1.83] |
OR/IRR | 1.07 | [0.97;1.2] | [1;1.16] |
OR hurdle | 1.36 | [1.02;1.87] | [1.13;1.66] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r21_continuous_maternalage.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
Like r1, but we use a 30-years-later cutoff year for our birth cohorts, relaxing our censoring requirements.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 21147)
## Samples: 6 chains, each with iter = 2000; warmup = 300; thin = 1;
## total post-warmup samples = 10200
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 5218)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.26 0.01 0.23 0.28 3594 1
## sd(hu_Intercept) 0.64 0.03 0.58 0.71 2700 1
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.90 0.30 1.32 2.49 433
## paternalage 0.05 0.04 -0.03 0.14 2701
## birth_cohort1670M1700 0.04 0.33 -0.60 0.68 500
## birth_cohort1700M1720 -0.02 0.31 -0.63 0.59 458
## birth_cohort1720M1760 -0.14 0.30 -0.72 0.43 422
## birth_cohort1760M1765 -0.10 0.30 -0.68 0.49 429
## birth_cohort1765M1770 -0.24 0.30 -0.82 0.33 424
## birth_cohort1770M1775 -0.25 0.30 -0.82 0.33 424
## birth_cohort1775M1780 -0.17 0.30 -0.74 0.40 427
## birth_cohort1780M1785 -0.25 0.30 -0.82 0.33 427
## birth_cohort1785M1790 -0.21 0.30 -0.79 0.37 425
## birth_cohort1790M1795 -0.18 0.30 -0.76 0.39 422
## birth_cohort1795M1800 -0.24 0.30 -0.81 0.34 424
## birth_cohort1800M1805 -0.24 0.30 -0.81 0.33 422
## birth_cohort1805M1810 -0.25 0.30 -0.82 0.33 421
## birth_cohort1810M1815 -0.22 0.30 -0.80 0.35 421
## birth_cohort1815M1820 -0.29 0.30 -0.87 0.28 424
## birth_cohort1820M1825 -0.34 0.30 -0.92 0.24 421
## birth_cohort1825M1830 -0.33 0.30 -0.91 0.24 422
## birth_cohort1830M1835 -0.33 0.30 -0.91 0.24 422
## birth_cohort1835M1840 -0.48 0.30 -1.05 0.09 423
## birth_cohort1840M1845 -0.63 0.30 -1.20 -0.05 423
## birth_cohort1845M1850 -0.97 0.30 -1.54 -0.38 428
## birth_cohort1850M1855 -0.63 0.30 -1.22 -0.04 430
## birth_cohort1855M1860 -0.26 0.30 -0.84 0.33 436
## birth_cohort1860M1865 -0.10 0.30 -0.69 0.48 430
## male1 0.08 0.01 0.05 0.11 10200
## maternalage.factor1420 -0.06 0.07 -0.21 0.07 10200
## maternalage.factor3550 -0.01 0.02 -0.06 0.03 10200
## paternalage.mean -0.07 0.05 -0.16 0.01 2718
## paternal_loss01 -0.16 0.07 -0.29 -0.02 10200
## paternal_loss15 -0.07 0.04 -0.15 0.02 4959
## paternal_loss510 -0.07 0.04 -0.15 0.00 4482
## paternal_loss1015 0.00 0.03 -0.07 0.07 4025
## paternal_loss1520 -0.10 0.03 -0.17 -0.04 3912
## paternal_loss2025 -0.12 0.03 -0.18 -0.06 3822
## paternal_loss2530 -0.03 0.03 -0.09 0.03 3697
## paternal_loss3035 -0.04 0.03 -0.10 0.02 3781
## paternal_loss3540 0.00 0.03 -0.06 0.05 4130
## paternal_loss4045 -0.01 0.03 -0.07 0.05 5370
## paternal_lossunclear -0.10 0.04 -0.17 -0.03 4179
## maternal_loss01 0.09 0.07 -0.04 0.22 10200
## maternal_loss15 -0.05 0.04 -0.14 0.03 5860
## maternal_loss510 0.02 0.04 -0.05 0.09 5452
## maternal_loss1015 -0.04 0.04 -0.11 0.03 5615
## maternal_loss1520 -0.02 0.03 -0.09 0.05 6052
## maternal_loss2025 -0.02 0.03 -0.09 0.04 5663
## maternal_loss2530 -0.07 0.03 -0.13 -0.01 5127
## maternal_loss3035 -0.04 0.03 -0.10 0.01 4720
## maternal_loss3540 -0.04 0.03 -0.09 0.01 4981
## maternal_loss4045 -0.02 0.03 -0.07 0.04 6097
## maternal_lossunclear -0.13 0.04 -0.20 -0.06 5711
## older_siblings1 0.00 0.02 -0.04 0.05 4953
## older_siblings2 -0.06 0.03 -0.12 0.00 3296
## older_siblings3 -0.08 0.04 -0.16 0.00 2852
## older_siblings4 -0.06 0.05 -0.16 0.03 2948
## older_siblings5P -0.07 0.06 -0.19 0.06 2779
## nr.siblings 0.00 0.01 -0.01 0.01 3260
## last_born1 -0.01 0.02 -0.05 0.02 10200
## hu_Intercept -0.98 0.71 -2.33 0.44 157
## hu_paternalage 0.29 0.11 0.08 0.51 2171
## hu_birth_cohort1670M1700 0.70 0.79 -0.89 2.26 196
## hu_birth_cohort1700M1720 -0.22 0.74 -1.67 1.18 176
## hu_birth_cohort1720M1760 0.47 0.70 -0.95 1.79 152
## hu_birth_cohort1760M1765 0.44 0.71 -1.00 1.79 156
## hu_birth_cohort1765M1770 0.24 0.70 -1.18 1.60 152
## hu_birth_cohort1770M1775 0.39 0.70 -1.01 1.73 153
## hu_birth_cohort1775M1780 0.32 0.70 -1.08 1.67 150
## hu_birth_cohort1780M1785 0.22 0.70 -1.21 1.55 153
## hu_birth_cohort1785M1790 0.08 0.70 -1.33 1.40 152
## hu_birth_cohort1790M1795 0.19 0.70 -1.23 1.52 153
## hu_birth_cohort1795M1800 -0.03 0.70 -1.43 1.30 150
## hu_birth_cohort1800M1805 -0.12 0.70 -1.54 1.21 151
## hu_birth_cohort1805M1810 0.16 0.70 -1.25 1.49 151
## hu_birth_cohort1810M1815 0.00 0.70 -1.41 1.33 150
## hu_birth_cohort1815M1820 -0.26 0.70 -1.68 1.07 151
## hu_birth_cohort1820M1825 -0.10 0.70 -1.50 1.22 151
## hu_birth_cohort1825M1830 -0.14 0.70 -1.55 1.18 150
## hu_birth_cohort1830M1835 -0.15 0.70 -1.56 1.17 151
## hu_birth_cohort1835M1840 0.01 0.70 -1.40 1.33 151
## hu_birth_cohort1840M1845 0.19 0.70 -1.23 1.50 151
## hu_birth_cohort1845M1850 0.73 0.70 -0.68 2.03 151
## hu_birth_cohort1850M1855 1.58 0.70 0.16 2.90 151
## hu_birth_cohort1855M1860 2.34 0.70 0.91 3.68 155
## hu_birth_cohort1860M1865 2.19 0.70 0.76 3.51 155
## hu_male1 0.35 0.04 0.28 0.42 10200
## hu_maternalage.factor1420 0.12 0.18 -0.22 0.48 10200
## hu_maternalage.factor3550 0.04 0.06 -0.07 0.15 10200
## hu_paternalage.mean -0.17 0.11 -0.39 0.06 2286
## hu_paternal_loss01 0.79 0.16 0.49 1.09 10200
## hu_paternal_loss15 0.59 0.11 0.37 0.81 4501
## hu_paternal_loss510 0.30 0.10 0.11 0.49 4148
## hu_paternal_loss1015 0.32 0.09 0.14 0.50 3686
## hu_paternal_loss1520 0.20 0.09 0.03 0.37 3892
## hu_paternal_loss2025 0.16 0.08 0.00 0.32 3524
## hu_paternal_loss2530 0.08 0.08 -0.08 0.25 3341
## hu_paternal_loss3035 0.02 0.08 -0.14 0.17 3759
## hu_paternal_loss3540 0.05 0.08 -0.11 0.20 3405
## hu_paternal_loss4045 0.04 0.09 -0.13 0.21 4719
## hu_paternal_lossunclear 0.74 0.09 0.57 0.92 3527
## hu_maternal_loss01 1.64 0.16 1.33 1.95 10200
## hu_maternal_loss15 0.77 0.10 0.57 0.98 5140
## hu_maternal_loss510 0.76 0.09 0.58 0.94 5361
## hu_maternal_loss1015 0.51 0.09 0.34 0.68 4913
## hu_maternal_loss1520 0.47 0.09 0.29 0.64 5080
## hu_maternal_loss2025 0.28 0.08 0.12 0.44 4974
## hu_maternal_loss2530 0.17 0.08 0.02 0.33 4773
## hu_maternal_loss3035 0.20 0.08 0.05 0.35 5380
## hu_maternal_loss3540 0.13 0.07 -0.01 0.28 5535
## hu_maternal_loss4045 0.29 0.08 0.14 0.44 6256
## hu_maternal_lossunclear 0.83 0.08 0.67 0.99 4622
## hu_older_siblings1 -0.01 0.06 -0.12 0.11 4497
## hu_older_siblings2 -0.08 0.07 -0.22 0.07 2490
## hu_older_siblings3 -0.13 0.10 -0.32 0.06 2217
## hu_older_siblings4 -0.13 0.12 -0.38 0.11 2010
## hu_older_siblings5P -0.39 0.16 -0.70 -0.07 1996
## hu_nr.siblings 0.08 0.01 0.05 0.11 2341
## hu_last_born1 0.02 0.05 -0.07 0.11 10200
## Rhat
## Intercept 1.02
## paternalage 1.00
## birth_cohort1670M1700 1.02
## birth_cohort1700M1720 1.02
## birth_cohort1720M1760 1.02
## birth_cohort1760M1765 1.02
## birth_cohort1765M1770 1.02
## birth_cohort1770M1775 1.02
## birth_cohort1775M1780 1.02
## birth_cohort1780M1785 1.02
## birth_cohort1785M1790 1.02
## birth_cohort1790M1795 1.02
## birth_cohort1795M1800 1.02
## birth_cohort1800M1805 1.02
## birth_cohort1805M1810 1.02
## birth_cohort1810M1815 1.02
## birth_cohort1815M1820 1.02
## birth_cohort1820M1825 1.02
## birth_cohort1825M1830 1.02
## birth_cohort1830M1835 1.02
## birth_cohort1835M1840 1.02
## birth_cohort1840M1845 1.02
## birth_cohort1845M1850 1.02
## birth_cohort1850M1855 1.02
## birth_cohort1855M1860 1.02
## birth_cohort1860M1865 1.02
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.00
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## paternal_lossunclear 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## maternal_lossunclear 1.00
## older_siblings1 1.00
## older_siblings2 1.00
## older_siblings3 1.00
## older_siblings4 1.00
## older_siblings5P 1.00
## nr.siblings 1.00
## last_born1 1.00
## hu_Intercept 1.04
## hu_paternalage 1.00
## hu_birth_cohort1670M1700 1.03
## hu_birth_cohort1700M1720 1.03
## hu_birth_cohort1720M1760 1.04
## hu_birth_cohort1760M1765 1.04
## hu_birth_cohort1765M1770 1.04
## hu_birth_cohort1770M1775 1.04
## hu_birth_cohort1775M1780 1.04
## hu_birth_cohort1780M1785 1.04
## hu_birth_cohort1785M1790 1.04
## hu_birth_cohort1790M1795 1.04
## hu_birth_cohort1795M1800 1.04
## hu_birth_cohort1800M1805 1.04
## hu_birth_cohort1805M1810 1.04
## hu_birth_cohort1810M1815 1.04
## hu_birth_cohort1815M1820 1.04
## hu_birth_cohort1820M1825 1.04
## hu_birth_cohort1825M1830 1.04
## hu_birth_cohort1830M1835 1.04
## hu_birth_cohort1835M1840 1.04
## hu_birth_cohort1840M1845 1.04
## hu_birth_cohort1845M1850 1.04
## hu_birth_cohort1850M1855 1.04
## hu_birth_cohort1855M1860 1.04
## hu_birth_cohort1860M1865 1.04
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.00
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_paternal_lossunclear 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_maternal_lossunclear 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.00
## hu_older_siblings3 1.00
## hu_older_siblings4 1.00
## hu_older_siblings5P 1.00
## hu_nr.siblings 1.00
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 6.69 | 3.735 | 12.02 |
paternalage | 1.054 | 0.9693 | 1.147 |
birth_cohort1670M1700 | 1.037 | 0.5473 | 1.968 |
birth_cohort1700M1720 | 0.9782 | 0.5322 | 1.798 |
birth_cohort1720M1760 | 0.8659 | 0.4846 | 1.534 |
birth_cohort1760M1765 | 0.9069 | 0.5051 | 1.632 |
birth_cohort1765M1770 | 0.7831 | 0.4384 | 1.396 |
birth_cohort1770M1775 | 0.7817 | 0.4397 | 1.397 |
birth_cohort1775M1780 | 0.8479 | 0.4761 | 1.499 |
birth_cohort1780M1785 | 0.7798 | 0.439 | 1.391 |
birth_cohort1785M1790 | 0.8127 | 0.4561 | 1.446 |
birth_cohort1790M1795 | 0.8312 | 0.4679 | 1.482 |
birth_cohort1795M1800 | 0.7896 | 0.4464 | 1.405 |
birth_cohort1800M1805 | 0.7847 | 0.4433 | 1.396 |
birth_cohort1805M1810 | 0.7827 | 0.4393 | 1.391 |
birth_cohort1810M1815 | 0.8006 | 0.4506 | 1.424 |
birth_cohort1815M1820 | 0.7454 | 0.4208 | 1.326 |
birth_cohort1820M1825 | 0.7108 | 0.3972 | 1.265 |
birth_cohort1825M1830 | 0.7156 | 0.4035 | 1.275 |
birth_cohort1830M1835 | 0.7168 | 0.4037 | 1.273 |
birth_cohort1835M1840 | 0.6207 | 0.3492 | 1.097 |
birth_cohort1840M1845 | 0.534 | 0.3016 | 0.9524 |
birth_cohort1845M1850 | 0.379 | 0.214 | 0.6854 |
birth_cohort1850M1855 | 0.5315 | 0.2947 | 0.9599 |
birth_cohort1855M1860 | 0.7715 | 0.431 | 1.391 |
birth_cohort1860M1865 | 0.9013 | 0.5034 | 1.619 |
male1 | 1.082 | 1.052 | 1.112 |
maternalage.factor1420 | 0.9375 | 0.8142 | 1.077 |
maternalage.factor3550 | 0.986 | 0.9424 | 1.032 |
paternalage.mean | 0.9283 | 0.8501 | 1.014 |
paternal_loss01 | 0.8562 | 0.752 | 0.9768 |
paternal_loss15 | 0.9365 | 0.8579 | 1.023 |
paternal_loss510 | 0.9288 | 0.863 | 1 |
paternal_loss1015 | 0.9998 | 0.9326 | 1.071 |
paternal_loss1520 | 0.9034 | 0.8444 | 0.9647 |
paternal_loss2025 | 0.8882 | 0.8326 | 0.9459 |
paternal_loss2530 | 0.966 | 0.9102 | 1.026 |
paternal_loss3035 | 0.961 | 0.9073 | 1.019 |
paternal_loss3540 | 0.9957 | 0.9407 | 1.054 |
paternal_loss4045 | 0.988 | 0.9293 | 1.051 |
paternal_lossunclear | 0.9038 | 0.8418 | 0.9707 |
maternal_loss01 | 1.096 | 0.9565 | 1.251 |
maternal_loss15 | 0.9499 | 0.8735 | 1.031 |
maternal_loss510 | 1.021 | 0.9466 | 1.098 |
maternal_loss1015 | 0.9652 | 0.8999 | 1.035 |
maternal_loss1520 | 0.9804 | 0.9156 | 1.049 |
maternal_loss2025 | 0.9789 | 0.9177 | 1.044 |
maternal_loss2530 | 0.9326 | 0.8783 | 0.9903 |
maternal_loss3035 | 0.9568 | 0.9046 | 1.013 |
maternal_loss3540 | 0.9614 | 0.9129 | 1.012 |
maternal_loss4045 | 0.982 | 0.929 | 1.037 |
maternal_lossunclear | 0.8769 | 0.8195 | 0.9397 |
older_siblings1 | 1.004 | 0.9611 | 1.049 |
older_siblings2 | 0.9418 | 0.8887 | 0.9988 |
older_siblings3 | 0.9242 | 0.8563 | 0.9974 |
older_siblings4 | 0.9395 | 0.8534 | 1.033 |
older_siblings5P | 0.9369 | 0.8255 | 1.063 |
nr.siblings | 1.002 | 0.99 | 1.014 |
last_born1 | 0.9864 | 0.9499 | 1.025 |
hu_Intercept | 0.3762 | 0.09703 | 1.555 |
hu_paternalage | 1.338 | 1.081 | 1.665 |
hu_birth_cohort1670M1700 | 2.005 | 0.4108 | 9.566 |
hu_birth_cohort1700M1720 | 0.7999 | 0.1875 | 3.239 |
hu_birth_cohort1720M1760 | 1.596 | 0.3877 | 6.019 |
hu_birth_cohort1760M1765 | 1.56 | 0.3692 | 6.007 |
hu_birth_cohort1765M1770 | 1.272 | 0.3062 | 4.933 |
hu_birth_cohort1770M1775 | 1.48 | 0.3644 | 5.619 |
hu_birth_cohort1775M1780 | 1.379 | 0.3395 | 5.29 |
hu_birth_cohort1780M1785 | 1.245 | 0.2984 | 4.725 |
hu_birth_cohort1785M1790 | 1.08 | 0.2632 | 4.063 |
hu_birth_cohort1790M1795 | 1.211 | 0.2936 | 4.566 |
hu_birth_cohort1795M1800 | 0.9722 | 0.2383 | 3.665 |
hu_birth_cohort1800M1805 | 0.8864 | 0.2151 | 3.345 |
hu_birth_cohort1805M1810 | 1.17 | 0.2866 | 4.422 |
hu_birth_cohort1810M1815 | 1.001 | 0.2432 | 3.772 |
hu_birth_cohort1815M1820 | 0.7746 | 0.186 | 2.924 |
hu_birth_cohort1820M1825 | 0.9092 | 0.2224 | 3.372 |
hu_birth_cohort1825M1830 | 0.8654 | 0.2122 | 3.262 |
hu_birth_cohort1830M1835 | 0.8595 | 0.2101 | 3.232 |
hu_birth_cohort1835M1840 | 1.011 | 0.2473 | 3.783 |
hu_birth_cohort1840M1845 | 1.205 | 0.2927 | 4.475 |
hu_birth_cohort1845M1850 | 2.066 | 0.5052 | 7.636 |
hu_birth_cohort1850M1855 | 4.859 | 1.175 | 18.21 |
hu_birth_cohort1855M1860 | 10.37 | 2.495 | 39.48 |
hu_birth_cohort1860M1865 | 8.909 | 2.143 | 33.32 |
hu_male1 | 1.417 | 1.322 | 1.517 |
hu_maternalage.factor1420 | 1.133 | 0.7995 | 1.617 |
hu_maternalage.factor3550 | 1.038 | 0.9317 | 1.159 |
hu_paternalage.mean | 0.847 | 0.6766 | 1.057 |
hu_paternal_loss01 | 2.199 | 1.627 | 2.972 |
hu_paternal_loss15 | 1.805 | 1.452 | 2.241 |
hu_paternal_loss510 | 1.354 | 1.12 | 1.63 |
hu_paternal_loss1015 | 1.382 | 1.154 | 1.653 |
hu_paternal_loss1520 | 1.217 | 1.027 | 1.444 |
hu_paternal_loss2025 | 1.176 | 0.9983 | 1.382 |
hu_paternal_loss2530 | 1.086 | 0.9271 | 1.279 |
hu_paternal_loss3035 | 1.021 | 0.87 | 1.191 |
hu_paternal_loss3540 | 1.047 | 0.8973 | 1.219 |
hu_paternal_loss4045 | 1.043 | 0.8812 | 1.236 |
hu_paternal_lossunclear | 2.1 | 1.772 | 2.498 |
hu_maternal_loss01 | 5.147 | 3.788 | 7.052 |
hu_maternal_loss15 | 2.163 | 1.768 | 2.654 |
hu_maternal_loss510 | 2.135 | 1.781 | 2.565 |
hu_maternal_loss1015 | 1.664 | 1.4 | 1.978 |
hu_maternal_loss1520 | 1.594 | 1.341 | 1.893 |
hu_maternal_loss2025 | 1.323 | 1.124 | 1.558 |
hu_maternal_loss2530 | 1.185 | 1.017 | 1.384 |
hu_maternal_loss3035 | 1.22 | 1.052 | 1.417 |
hu_maternal_loss3540 | 1.14 | 0.9909 | 1.317 |
hu_maternal_loss4045 | 1.337 | 1.152 | 1.546 |
hu_maternal_lossunclear | 2.293 | 1.951 | 2.685 |
hu_older_siblings1 | 0.9922 | 0.8875 | 1.111 |
hu_older_siblings2 | 0.9263 | 0.7992 | 1.071 |
hu_older_siblings3 | 0.8774 | 0.725 | 1.06 |
hu_older_siblings4 | 0.8749 | 0.6868 | 1.112 |
hu_older_siblings5P | 0.6798 | 0.4945 | 0.9293 |
hu_nr.siblings | 1.084 | 1.054 | 1.117 |
hu_last_born1 | 1.019 | 0.93 | 1.118 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 3.52 | [1.4;7.02] | [1.95;5.74] |
estimate father 35y | 3.29 | [1.16;6.89] | [1.71;5.55] |
percentage change | -6.76 | [-23.07;6.41] | [-17.23;2.28] |
OR/IRR | 1.05 | [0.97;1.15] | [1;1.11] |
OR hurdle | 1.34 | [1.08;1.66] | [1.16;1.55] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r22_relaxed_exclusion_censoring.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
To demonstrate the robustness of our prior choice we use Student’s t priors (fatter tails than normal priors) for our population-level effects and a half-Cauchy prior for our group-level effect for the family.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ student_t(5,0,10)
## sd ~ cauchy(0,5)
## b_hu ~ student_t(5,0,10)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.23 0.01 0.20 0.25 1097 1.00
## sd(hu_Intercept) 0.48 0.04 0.39 0.56 913 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.65 0.08 1.50 1.80 1626
## paternalage 0.07 0.05 -0.04 0.17 896
## birth_cohort1760M1765 0.00 0.06 -0.14 0.12 3000
## birth_cohort1765M1770 -0.12 0.06 -0.23 0.00 1296
## birth_cohort1770M1775 -0.11 0.06 -0.23 0.00 1119
## birth_cohort1775M1780 -0.03 0.06 -0.13 0.09 1038
## birth_cohort1780M1785 -0.11 0.06 -0.22 0.01 1144
## birth_cohort1785M1790 -0.10 0.05 -0.20 0.01 1024
## birth_cohort1790M1795 -0.08 0.05 -0.18 0.03 903
## birth_cohort1795M1800 -0.11 0.05 -0.20 -0.01 838
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.02 813
## birth_cohort1805M1810 -0.14 0.05 -0.23 -0.04 881
## birth_cohort1810M1815 -0.10 0.05 -0.19 0.00 758
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 718
## birth_cohort1820M1825 -0.19 0.05 -0.28 -0.09 772
## birth_cohort1825M1830 -0.21 0.05 -0.30 -0.12 683
## birth_cohort1830M1835 -0.18 0.05 -0.28 -0.09 810
## male1 0.08 0.02 0.04 0.11 3000
## maternalage.factor1420 -0.05 0.09 -0.24 0.14 3000
## maternalage.factor3550 0.00 0.03 -0.06 0.05 3000
## paternalage.mean -0.07 0.05 -0.18 0.04 902
## paternal_loss01 -0.15 0.07 -0.30 -0.01 3000
## paternal_loss15 -0.04 0.05 -0.13 0.06 3000
## paternal_loss510 -0.07 0.04 -0.15 0.02 1493
## paternal_loss1015 0.01 0.04 -0.07 0.08 1440
## paternal_loss1520 -0.09 0.04 -0.17 -0.02 1270
## paternal_loss2025 -0.12 0.04 -0.19 -0.04 1360
## paternal_loss2530 -0.01 0.03 -0.08 0.06 1287
## paternal_loss3035 -0.03 0.03 -0.09 0.04 1453
## paternal_loss3540 -0.01 0.03 -0.07 0.06 1482
## paternal_loss4045 -0.01 0.04 -0.08 0.06 3000
## maternal_loss01 0.10 0.08 -0.05 0.26 3000
## maternal_loss15 -0.02 0.05 -0.11 0.07 3000
## maternal_loss510 0.07 0.04 -0.01 0.15 3000
## maternal_loss1015 0.03 0.04 -0.06 0.11 3000
## maternal_loss1520 0.00 0.04 -0.08 0.08 3000
## maternal_loss2025 0.00 0.04 -0.07 0.08 3000
## maternal_loss2530 -0.02 0.03 -0.09 0.05 3000
## maternal_loss3035 -0.05 0.03 -0.11 0.01 3000
## maternal_loss3540 -0.03 0.03 -0.09 0.03 3000
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## older_siblings1 0.03 0.03 -0.02 0.08 1604
## older_siblings2 -0.05 0.04 -0.12 0.02 1083
## older_siblings3 -0.07 0.05 -0.16 0.02 930
## older_siblings4 -0.09 0.06 -0.21 0.02 914
## older_siblings5P -0.09 0.08 -0.24 0.05 900
## nr.siblings 0.01 0.01 0.00 0.02 1167
## last_born1 -0.04 0.02 -0.09 0.00 3000
## hu_Intercept -0.33 0.19 -0.69 0.06 1531
## hu_paternalage 0.26 0.14 -0.02 0.54 818
## hu_birth_cohort1760M1765 -0.05 0.17 -0.38 0.30 3000
## hu_birth_cohort1765M1770 -0.31 0.15 -0.60 -0.01 1359
## hu_birth_cohort1770M1775 -0.07 0.14 -0.34 0.20 1221
## hu_birth_cohort1775M1780 -0.20 0.14 -0.47 0.08 1111
## hu_birth_cohort1780M1785 -0.28 0.15 -0.56 0.00 1160
## hu_birth_cohort1785M1790 -0.42 0.14 -0.69 -0.14 1184
## hu_birth_cohort1790M1795 -0.31 0.14 -0.58 -0.06 1151
## hu_birth_cohort1795M1800 -0.46 0.13 -0.71 -0.21 956
## hu_birth_cohort1800M1805 -0.54 0.13 -0.78 -0.29 1068
## hu_birth_cohort1805M1810 -0.27 0.13 -0.51 -0.02 984
## hu_birth_cohort1810M1815 -0.44 0.13 -0.69 -0.19 951
## hu_birth_cohort1815M1820 -0.70 0.12 -0.93 -0.47 889
## hu_birth_cohort1820M1825 -0.52 0.12 -0.75 -0.29 858
## hu_birth_cohort1825M1830 -0.54 0.12 -0.77 -0.31 869
## hu_birth_cohort1830M1835 -0.55 0.12 -0.79 -0.32 947
## hu_male1 0.27 0.04 0.19 0.36 3000
## hu_maternalage.factor1420 0.24 0.24 -0.22 0.71 3000
## hu_maternalage.factor3550 0.13 0.07 -0.01 0.28 3000
## hu_paternalage.mean -0.18 0.15 -0.47 0.11 832
## hu_paternal_loss01 0.57 0.18 0.20 0.93 3000
## hu_paternal_loss15 0.53 0.13 0.27 0.79 3000
## hu_paternal_loss510 0.19 0.11 -0.02 0.42 3000
## hu_paternal_loss1015 0.16 0.10 -0.04 0.36 2119
## hu_paternal_loss1520 0.10 0.10 -0.09 0.30 1962
## hu_paternal_loss2025 0.15 0.10 -0.04 0.34 2042
## hu_paternal_loss2530 0.06 0.09 -0.12 0.24 1828
## hu_paternal_loss3035 -0.02 0.09 -0.21 0.15 1968
## hu_paternal_loss3540 -0.02 0.09 -0.18 0.16 2103
## hu_paternal_loss4045 0.14 0.10 -0.05 0.34 3000
## hu_maternal_loss01 1.58 0.19 1.22 1.97 3000
## hu_maternal_loss15 0.58 0.12 0.35 0.82 3000
## hu_maternal_loss510 0.48 0.11 0.27 0.70 3000
## hu_maternal_loss1015 0.47 0.11 0.25 0.69 3000
## hu_maternal_loss1520 0.31 0.11 0.10 0.53 3000
## hu_maternal_loss2025 0.26 0.10 0.05 0.46 3000
## hu_maternal_loss2530 0.19 0.09 0.01 0.37 3000
## hu_maternal_loss3035 0.22 0.09 0.05 0.39 3000
## hu_maternal_loss3540 0.07 0.08 -0.09 0.22 3000
## hu_maternal_loss4045 0.28 0.09 0.11 0.45 3000
## hu_older_siblings1 -0.01 0.07 -0.15 0.14 3000
## hu_older_siblings2 -0.15 0.10 -0.34 0.04 1077
## hu_older_siblings3 -0.19 0.13 -0.43 0.06 905
## hu_older_siblings4 -0.22 0.16 -0.54 0.09 873
## hu_older_siblings5P -0.56 0.21 -0.96 -0.15 799
## hu_nr.siblings 0.10 0.02 0.07 0.14 1027
## hu_last_born1 0.08 0.06 -0.04 0.20 3000
## Rhat
## Intercept 1.00
## paternalage 1.00
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.00
## birth_cohort1770M1775 1.00
## birth_cohort1775M1780 1.01
## birth_cohort1780M1785 1.00
## birth_cohort1785M1790 1.00
## birth_cohort1790M1795 1.00
## birth_cohort1795M1800 1.00
## birth_cohort1800M1805 1.00
## birth_cohort1805M1810 1.00
## birth_cohort1810M1815 1.01
## birth_cohort1815M1820 1.00
## birth_cohort1820M1825 1.00
## birth_cohort1825M1830 1.01
## birth_cohort1830M1835 1.00
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.01
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.00
## older_siblings3 1.00
## older_siblings4 1.00
## older_siblings5P 1.00
## nr.siblings 1.00
## last_born1 1.00
## hu_Intercept 1.00
## hu_paternalage 1.00
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.00
## hu_birth_cohort1775M1780 1.00
## hu_birth_cohort1780M1785 1.00
## hu_birth_cohort1785M1790 1.00
## hu_birth_cohort1790M1795 1.00
## hu_birth_cohort1795M1800 1.00
## hu_birth_cohort1800M1805 1.00
## hu_birth_cohort1805M1810 1.00
## hu_birth_cohort1810M1815 1.01
## hu_birth_cohort1815M1820 1.00
## hu_birth_cohort1820M1825 1.00
## hu_birth_cohort1825M1830 1.01
## hu_birth_cohort1830M1835 1.00
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.00
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.00
## hu_older_siblings3 1.00
## hu_older_siblings4 1.00
## hu_older_siblings5P 1.00
## hu_nr.siblings 1.00
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.223 | 4.491 | 6.054 |
paternalage | 1.068 | 0.9598 | 1.183 |
birth_cohort1760M1765 | 0.9958 | 0.8734 | 1.126 |
birth_cohort1765M1770 | 0.8878 | 0.7917 | 0.9967 |
birth_cohort1770M1775 | 0.8914 | 0.7959 | 1.003 |
birth_cohort1775M1780 | 0.9749 | 0.8757 | 1.09 |
birth_cohort1780M1785 | 0.8971 | 0.8028 | 1.009 |
birth_cohort1785M1790 | 0.9092 | 0.8175 | 1.015 |
birth_cohort1790M1795 | 0.9276 | 0.8375 | 1.034 |
birth_cohort1795M1800 | 0.8996 | 0.8162 | 0.9918 |
birth_cohort1800M1805 | 0.8933 | 0.81 | 0.9837 |
birth_cohort1805M1810 | 0.873 | 0.7914 | 0.9617 |
birth_cohort1810M1815 | 0.9077 | 0.8278 | 1.002 |
birth_cohort1815M1820 | 0.8677 | 0.7946 | 0.9491 |
birth_cohort1820M1825 | 0.8311 | 0.7581 | 0.9134 |
birth_cohort1825M1830 | 0.8105 | 0.7385 | 0.8884 |
birth_cohort1830M1835 | 0.8333 | 0.7563 | 0.9173 |
male1 | 1.081 | 1.046 | 1.116 |
maternalage.factor1420 | 0.9483 | 0.7875 | 1.146 |
maternalage.factor3550 | 0.997 | 0.9441 | 1.05 |
paternalage.mean | 0.9336 | 0.8381 | 1.04 |
paternal_loss01 | 0.8574 | 0.7434 | 0.9861 |
paternal_loss15 | 0.9642 | 0.8763 | 1.066 |
paternal_loss510 | 0.9349 | 0.8602 | 1.017 |
paternal_loss1015 | 1.007 | 0.9322 | 1.087 |
paternal_loss1520 | 0.9115 | 0.8437 | 0.9831 |
paternal_loss2025 | 0.8882 | 0.8262 | 0.9582 |
paternal_loss2530 | 0.9921 | 0.9262 | 1.06 |
paternal_loss3035 | 0.9734 | 0.9112 | 1.038 |
paternal_loss3540 | 0.9917 | 0.93 | 1.059 |
paternal_loss4045 | 0.9916 | 0.9223 | 1.065 |
maternal_loss01 | 1.106 | 0.9499 | 1.296 |
maternal_loss15 | 0.9801 | 0.8927 | 1.073 |
maternal_loss510 | 1.072 | 0.9908 | 1.161 |
maternal_loss1015 | 1.027 | 0.9422 | 1.116 |
maternal_loss1520 | 1.003 | 0.9265 | 1.085 |
maternal_loss2025 | 1.005 | 0.9309 | 1.08 |
maternal_loss2530 | 0.9792 | 0.9151 | 1.049 |
maternal_loss3035 | 0.9503 | 0.8934 | 1.014 |
maternal_loss3540 | 0.9675 | 0.9131 | 1.027 |
maternal_loss4045 | 0.9716 | 0.9101 | 1.033 |
older_siblings1 | 1.027 | 0.9764 | 1.081 |
older_siblings2 | 0.9542 | 0.8897 | 1.024 |
older_siblings3 | 0.9302 | 0.8506 | 1.02 |
older_siblings4 | 0.9121 | 0.8113 | 1.024 |
older_siblings5P | 0.91 | 0.7833 | 1.056 |
nr.siblings | 1.01 | 0.996 | 1.024 |
last_born1 | 0.9576 | 0.913 | 1.003 |
hu_Intercept | 0.7201 | 0.4994 | 1.064 |
hu_paternalage | 1.294 | 0.9804 | 1.719 |
hu_birth_cohort1760M1765 | 0.9508 | 0.6854 | 1.345 |
hu_birth_cohort1765M1770 | 0.7335 | 0.549 | 0.9923 |
hu_birth_cohort1770M1775 | 0.9363 | 0.7122 | 1.218 |
hu_birth_cohort1775M1780 | 0.8198 | 0.628 | 1.084 |
hu_birth_cohort1780M1785 | 0.757 | 0.571 | 1.004 |
hu_birth_cohort1785M1790 | 0.6582 | 0.5038 | 0.8708 |
hu_birth_cohort1790M1795 | 0.7319 | 0.5606 | 0.9449 |
hu_birth_cohort1795M1800 | 0.6322 | 0.4902 | 0.8077 |
hu_birth_cohort1800M1805 | 0.5847 | 0.4572 | 0.7452 |
hu_birth_cohort1805M1810 | 0.7654 | 0.5991 | 0.9836 |
hu_birth_cohort1810M1815 | 0.6461 | 0.5019 | 0.8251 |
hu_birth_cohort1815M1820 | 0.4964 | 0.3938 | 0.6253 |
hu_birth_cohort1820M1825 | 0.5926 | 0.4735 | 0.7519 |
hu_birth_cohort1825M1830 | 0.5807 | 0.4615 | 0.7338 |
hu_birth_cohort1830M1835 | 0.575 | 0.4556 | 0.7244 |
hu_male1 | 1.315 | 1.206 | 1.427 |
hu_maternalage.factor1420 | 1.274 | 0.8023 | 2.04 |
hu_maternalage.factor3550 | 1.142 | 0.9905 | 1.318 |
hu_paternalage.mean | 0.8382 | 0.6277 | 1.121 |
hu_paternal_loss01 | 1.774 | 1.225 | 2.534 |
hu_paternal_loss15 | 1.698 | 1.314 | 2.209 |
hu_paternal_loss510 | 1.214 | 0.9782 | 1.519 |
hu_paternal_loss1015 | 1.17 | 0.9578 | 1.433 |
hu_paternal_loss1520 | 1.108 | 0.9104 | 1.345 |
hu_paternal_loss2025 | 1.166 | 0.9644 | 1.407 |
hu_paternal_loss2530 | 1.061 | 0.8878 | 1.269 |
hu_paternal_loss3035 | 0.9757 | 0.8079 | 1.163 |
hu_paternal_loss3540 | 0.9845 | 0.8348 | 1.169 |
hu_paternal_loss4045 | 1.153 | 0.9557 | 1.411 |
hu_maternal_loss01 | 4.87 | 3.392 | 7.143 |
hu_maternal_loss15 | 1.791 | 1.414 | 2.275 |
hu_maternal_loss510 | 1.621 | 1.314 | 2.008 |
hu_maternal_loss1015 | 1.601 | 1.288 | 1.988 |
hu_maternal_loss1520 | 1.364 | 1.101 | 1.704 |
hu_maternal_loss2025 | 1.295 | 1.055 | 1.592 |
hu_maternal_loss2530 | 1.208 | 1.009 | 1.447 |
hu_maternal_loss3035 | 1.242 | 1.049 | 1.483 |
hu_maternal_loss3540 | 1.069 | 0.9138 | 1.25 |
hu_maternal_loss4045 | 1.32 | 1.121 | 1.572 |
hu_older_siblings1 | 0.9909 | 0.8571 | 1.149 |
hu_older_siblings2 | 0.8589 | 0.7084 | 1.043 |
hu_older_siblings3 | 0.8269 | 0.6502 | 1.066 |
hu_older_siblings4 | 0.7993 | 0.5854 | 1.09 |
hu_older_siblings5P | 0.5689 | 0.3823 | 0.8605 |
hu_nr.siblings | 1.109 | 1.07 | 1.149 |
hu_last_born1 | 1.085 | 0.9574 | 1.226 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.28 | [1.92;2.68] | [2.04;2.53] |
estimate father 35y | 2.09 | [1.66;2.6] | [1.8;2.4] |
percentage change | -8.39 | [-25.46;10.78] | [-19.66;4.02] |
OR/IRR | 1.07 | [0.96;1.18] | [1;1.14] |
OR hurdle | 1.29 | [0.98;1.72] | [1.08;1.56] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r23_student_cauchy_priors.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
To demonstrate the robustness of our prior choice we use improper flat priors. These priors should make the model’s results comparable to a frequentist maximum likelihood approach.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + birth_cohort + male + maternalage.factor + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 800; warmup = 300; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## sd ~ student_t(3, 0, 10)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.23 0.01 0.20 0.25 1227 1.00
## sd(hu_Intercept) 0.47 0.05 0.38 0.56 766 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 1.65 0.08 1.50 1.81 1520
## paternalage 0.07 0.05 -0.04 0.18 906
## birth_cohort1760M1765 0.00 0.07 -0.13 0.13 3000
## birth_cohort1765M1770 -0.12 0.06 -0.23 0.00 1281
## birth_cohort1770M1775 -0.11 0.06 -0.23 0.00 1097
## birth_cohort1775M1780 -0.03 0.05 -0.13 0.08 1124
## birth_cohort1780M1785 -0.11 0.06 -0.22 0.01 1046
## birth_cohort1785M1790 -0.09 0.06 -0.20 0.01 870
## birth_cohort1790M1795 -0.08 0.05 -0.18 0.03 861
## birth_cohort1795M1800 -0.11 0.05 -0.20 -0.01 859
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.01 786
## birth_cohort1805M1810 -0.14 0.05 -0.23 -0.04 855
## birth_cohort1810M1815 -0.10 0.05 -0.19 0.00 797
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 755
## birth_cohort1820M1825 -0.18 0.05 -0.27 -0.09 820
## birth_cohort1825M1830 -0.21 0.05 -0.30 -0.12 763
## birth_cohort1830M1835 -0.18 0.05 -0.27 -0.08 705
## male1 0.08 0.02 0.04 0.11 3000
## maternalage.factor1420 -0.05 0.09 -0.24 0.13 3000
## maternalage.factor3550 0.00 0.03 -0.06 0.05 3000
## paternalage.mean -0.07 0.06 -0.18 0.04 946
## paternal_loss01 -0.15 0.07 -0.29 -0.01 3000
## paternal_loss15 -0.03 0.05 -0.13 0.06 1815
## paternal_loss510 -0.07 0.04 -0.15 0.02 1557
## paternal_loss1015 0.01 0.04 -0.07 0.09 1552
## paternal_loss1520 -0.09 0.04 -0.17 -0.02 1582
## paternal_loss2025 -0.12 0.04 -0.19 -0.04 1772
## paternal_loss2530 -0.01 0.04 -0.08 0.06 1427
## paternal_loss3035 -0.03 0.03 -0.09 0.04 1391
## paternal_loss3540 -0.01 0.03 -0.07 0.05 1505
## paternal_loss4045 -0.01 0.04 -0.08 0.06 3000
## maternal_loss01 0.10 0.08 -0.06 0.25 3000
## maternal_loss15 -0.02 0.05 -0.11 0.07 3000
## maternal_loss510 0.07 0.04 -0.01 0.15 3000
## maternal_loss1015 0.03 0.04 -0.06 0.11 3000
## maternal_loss1520 0.00 0.04 -0.08 0.09 3000
## maternal_loss2025 0.00 0.04 -0.07 0.08 3000
## maternal_loss2530 -0.02 0.04 -0.09 0.04 2292
## maternal_loss3035 -0.05 0.03 -0.12 0.01 2460
## maternal_loss3540 -0.03 0.03 -0.09 0.02 2191
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## older_siblings1 0.03 0.03 -0.03 0.08 3000
## older_siblings2 -0.05 0.04 -0.12 0.03 1064
## older_siblings3 -0.07 0.05 -0.17 0.02 961
## older_siblings4 -0.09 0.06 -0.21 0.03 940
## older_siblings5P -0.09 0.08 -0.25 0.06 938
## nr.siblings 0.01 0.01 0.00 0.02 1098
## last_born1 -0.04 0.02 -0.09 0.00 3000
## hu_Intercept -0.32 0.20 -0.72 0.07 1079
## hu_paternalage 0.27 0.15 -0.03 0.56 856
## hu_birth_cohort1760M1765 -0.06 0.17 -0.38 0.28 3000
## hu_birth_cohort1765M1770 -0.32 0.15 -0.61 -0.03 1133
## hu_birth_cohort1770M1775 -0.08 0.14 -0.36 0.19 1123
## hu_birth_cohort1775M1780 -0.21 0.14 -0.48 0.07 1168
## hu_birth_cohort1780M1785 -0.29 0.15 -0.59 0.00 1171
## hu_birth_cohort1785M1790 -0.43 0.15 -0.71 -0.15 1077
## hu_birth_cohort1790M1795 -0.33 0.14 -0.59 -0.07 1045
## hu_birth_cohort1795M1800 -0.47 0.13 -0.72 -0.22 966
## hu_birth_cohort1800M1805 -0.55 0.13 -0.80 -0.31 914
## hu_birth_cohort1805M1810 -0.28 0.13 -0.54 -0.04 898
## hu_birth_cohort1810M1815 -0.45 0.12 -0.70 -0.20 869
## hu_birth_cohort1815M1820 -0.72 0.12 -0.95 -0.48 806
## hu_birth_cohort1820M1825 -0.54 0.12 -0.76 -0.31 793
## hu_birth_cohort1825M1830 -0.56 0.12 -0.79 -0.33 834
## hu_birth_cohort1830M1835 -0.57 0.12 -0.81 -0.35 802
## hu_male1 0.27 0.04 0.19 0.36 3000
## hu_maternalage.factor1420 0.24 0.23 -0.22 0.70 3000
## hu_maternalage.factor3550 0.13 0.07 -0.01 0.27 3000
## hu_paternalage.mean -0.18 0.15 -0.49 0.12 855
## hu_paternal_loss01 0.57 0.18 0.23 0.93 3000
## hu_paternal_loss15 0.53 0.13 0.28 0.77 3000
## hu_paternal_loss510 0.19 0.11 -0.03 0.41 1767
## hu_paternal_loss1015 0.16 0.11 -0.05 0.36 1603
## hu_paternal_loss1520 0.10 0.10 -0.09 0.30 1572
## hu_paternal_loss2025 0.15 0.10 -0.04 0.34 1825
## hu_paternal_loss2530 0.06 0.09 -0.13 0.24 1456
## hu_paternal_loss3035 -0.02 0.09 -0.21 0.15 1577
## hu_paternal_loss3540 -0.01 0.09 -0.20 0.16 1553
## hu_paternal_loss4045 0.14 0.10 -0.06 0.34 3000
## hu_maternal_loss01 1.58 0.19 1.21 1.97 3000
## hu_maternal_loss15 0.59 0.12 0.35 0.83 3000
## hu_maternal_loss510 0.49 0.11 0.28 0.70 3000
## hu_maternal_loss1015 0.48 0.11 0.26 0.69 3000
## hu_maternal_loss1520 0.32 0.11 0.10 0.53 3000
## hu_maternal_loss2025 0.27 0.10 0.07 0.46 3000
## hu_maternal_loss2530 0.19 0.10 0.01 0.39 3000
## hu_maternal_loss3035 0.22 0.09 0.05 0.39 3000
## hu_maternal_loss3540 0.07 0.08 -0.09 0.23 3000
## hu_maternal_loss4045 0.28 0.09 0.12 0.45 3000
## hu_older_siblings1 -0.01 0.08 -0.16 0.14 1559
## hu_older_siblings2 -0.16 0.10 -0.36 0.04 940
## hu_older_siblings3 -0.20 0.13 -0.45 0.04 859
## hu_older_siblings4 -0.23 0.16 -0.55 0.09 829
## hu_older_siblings5P -0.58 0.22 -1.01 -0.16 786
## hu_nr.siblings 0.10 0.02 0.07 0.14 1018
## hu_last_born1 0.08 0.06 -0.04 0.20 3000
## Rhat
## Intercept 1.00
## paternalage 1.01
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.01
## birth_cohort1770M1775 1.00
## birth_cohort1775M1780 1.00
## birth_cohort1780M1785 1.00
## birth_cohort1785M1790 1.00
## birth_cohort1790M1795 1.00
## birth_cohort1795M1800 1.00
## birth_cohort1800M1805 1.01
## birth_cohort1805M1810 1.01
## birth_cohort1810M1815 1.01
## birth_cohort1815M1820 1.01
## birth_cohort1820M1825 1.00
## birth_cohort1825M1830 1.01
## birth_cohort1830M1835 1.00
## male1 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## paternalage.mean 1.01
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.01
## older_siblings3 1.01
## older_siblings4 1.01
## older_siblings5P 1.01
## nr.siblings 1.00
## last_born1 1.00
## hu_Intercept 1.00
## hu_paternalage 1.01
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.00
## hu_birth_cohort1775M1780 1.00
## hu_birth_cohort1780M1785 1.00
## hu_birth_cohort1785M1790 1.00
## hu_birth_cohort1790M1795 1.00
## hu_birth_cohort1795M1800 1.00
## hu_birth_cohort1800M1805 1.00
## hu_birth_cohort1805M1810 1.00
## hu_birth_cohort1810M1815 1.00
## hu_birth_cohort1815M1820 1.00
## hu_birth_cohort1820M1825 1.00
## hu_birth_cohort1825M1830 1.00
## hu_birth_cohort1830M1835 1.00
## hu_male1 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_paternalage.mean 1.01
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.01
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.01
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.00
## hu_older_siblings3 1.01
## hu_older_siblings4 1.01
## hu_older_siblings5P 1.01
## hu_nr.siblings 1.00
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 5.227 | 4.469 | 6.102 |
paternalage | 1.067 | 0.9629 | 1.197 |
birth_cohort1760M1765 | 0.9957 | 0.8777 | 1.139 |
birth_cohort1765M1770 | 0.8883 | 0.7951 | 0.9953 |
birth_cohort1770M1775 | 0.8916 | 0.7952 | 1 |
birth_cohort1775M1780 | 0.9751 | 0.8754 | 1.087 |
birth_cohort1780M1785 | 0.8963 | 0.8026 | 1.006 |
birth_cohort1785M1790 | 0.9098 | 0.818 | 1.013 |
birth_cohort1790M1795 | 0.9269 | 0.837 | 1.029 |
birth_cohort1795M1800 | 0.8997 | 0.8169 | 0.9932 |
birth_cohort1800M1805 | 0.8939 | 0.8137 | 0.9859 |
birth_cohort1805M1810 | 0.8737 | 0.7907 | 0.9637 |
birth_cohort1810M1815 | 0.9085 | 0.8269 | 0.9959 |
birth_cohort1815M1820 | 0.8686 | 0.7908 | 0.9523 |
birth_cohort1820M1825 | 0.8321 | 0.7609 | 0.9122 |
birth_cohort1825M1830 | 0.8112 | 0.7418 | 0.8909 |
birth_cohort1830M1835 | 0.8345 | 0.7612 | 0.9195 |
male1 | 1.081 | 1.045 | 1.118 |
maternalage.factor1420 | 0.9477 | 0.7873 | 1.137 |
maternalage.factor3550 | 0.9965 | 0.9425 | 1.05 |
paternalage.mean | 0.934 | 0.8345 | 1.041 |
paternal_loss01 | 0.859 | 0.7454 | 0.9939 |
paternal_loss15 | 0.966 | 0.874 | 1.065 |
paternal_loss510 | 0.9357 | 0.861 | 1.018 |
paternal_loss1015 | 1.008 | 0.9287 | 1.091 |
paternal_loss1520 | 0.9115 | 0.8472 | 0.9804 |
paternal_loss2025 | 0.888 | 0.8255 | 0.957 |
paternal_loss2530 | 0.9916 | 0.9254 | 1.064 |
paternal_loss3035 | 0.9733 | 0.9107 | 1.038 |
paternal_loss3540 | 0.9922 | 0.9325 | 1.056 |
paternal_loss4045 | 0.9915 | 0.923 | 1.066 |
maternal_loss01 | 1.106 | 0.9446 | 1.289 |
maternal_loss15 | 0.9807 | 0.8964 | 1.076 |
maternal_loss510 | 1.073 | 0.9909 | 1.162 |
maternal_loss1015 | 1.028 | 0.9428 | 1.113 |
maternal_loss1520 | 1.004 | 0.9217 | 1.091 |
maternal_loss2025 | 1.004 | 0.931 | 1.086 |
maternal_loss2530 | 0.9785 | 0.9119 | 1.046 |
maternal_loss3035 | 0.95 | 0.8911 | 1.015 |
maternal_loss3540 | 0.9672 | 0.9136 | 1.024 |
maternal_loss4045 | 0.9707 | 0.9128 | 1.035 |
older_siblings1 | 1.027 | 0.9722 | 1.083 |
older_siblings2 | 0.9537 | 0.8849 | 1.025 |
older_siblings3 | 0.9306 | 0.8457 | 1.021 |
older_siblings4 | 0.9142 | 0.81 | 1.03 |
older_siblings5P | 0.9105 | 0.7764 | 1.061 |
nr.siblings | 1.01 | 0.9952 | 1.024 |
last_born1 | 0.9573 | 0.9127 | 1.003 |
hu_Intercept | 0.7249 | 0.4844 | 1.068 |
hu_paternalage | 1.305 | 0.9704 | 1.754 |
hu_birth_cohort1760M1765 | 0.9403 | 0.6808 | 1.326 |
hu_birth_cohort1765M1770 | 0.7245 | 0.5449 | 0.9752 |
hu_birth_cohort1770M1775 | 0.9241 | 0.697 | 1.208 |
hu_birth_cohort1775M1780 | 0.8112 | 0.6157 | 1.076 |
hu_birth_cohort1780M1785 | 0.7486 | 0.5555 | 0.9974 |
hu_birth_cohort1785M1790 | 0.6495 | 0.49 | 0.8636 |
hu_birth_cohort1790M1795 | 0.7216 | 0.5522 | 0.9369 |
hu_birth_cohort1795M1800 | 0.6246 | 0.486 | 0.8028 |
hu_birth_cohort1800M1805 | 0.5759 | 0.4479 | 0.7352 |
hu_birth_cohort1805M1810 | 0.7541 | 0.5855 | 0.957 |
hu_birth_cohort1810M1815 | 0.6382 | 0.4985 | 0.8176 |
hu_birth_cohort1815M1820 | 0.4892 | 0.3852 | 0.617 |
hu_birth_cohort1820M1825 | 0.5846 | 0.4655 | 0.7328 |
hu_birth_cohort1825M1830 | 0.5736 | 0.4517 | 0.722 |
hu_birth_cohort1830M1835 | 0.5667 | 0.4452 | 0.7076 |
hu_male1 | 1.315 | 1.204 | 1.435 |
hu_maternalage.factor1420 | 1.265 | 0.8 | 2.007 |
hu_maternalage.factor3550 | 1.138 | 0.9871 | 1.314 |
hu_paternalage.mean | 0.8324 | 0.6145 | 1.126 |
hu_paternal_loss01 | 1.769 | 1.255 | 2.522 |
hu_paternal_loss15 | 1.696 | 1.322 | 2.17 |
hu_paternal_loss510 | 1.212 | 0.9715 | 1.504 |
hu_paternal_loss1015 | 1.171 | 0.9505 | 1.435 |
hu_paternal_loss1520 | 1.109 | 0.9095 | 1.347 |
hu_paternal_loss2025 | 1.165 | 0.9627 | 1.402 |
hu_paternal_loss2530 | 1.064 | 0.8823 | 1.273 |
hu_paternal_loss3035 | 0.9774 | 0.8131 | 1.165 |
hu_paternal_loss3540 | 0.9856 | 0.8219 | 1.176 |
hu_paternal_loss4045 | 1.154 | 0.9458 | 1.408 |
hu_maternal_loss01 | 4.867 | 3.366 | 7.175 |
hu_maternal_loss15 | 1.801 | 1.423 | 2.288 |
hu_maternal_loss510 | 1.625 | 1.319 | 2.008 |
hu_maternal_loss1015 | 1.609 | 1.3 | 1.989 |
hu_maternal_loss1520 | 1.374 | 1.108 | 1.707 |
hu_maternal_loss2025 | 1.305 | 1.069 | 1.589 |
hu_maternal_loss2530 | 1.215 | 1.007 | 1.471 |
hu_maternal_loss3035 | 1.246 | 1.052 | 1.481 |
hu_maternal_loss3540 | 1.074 | 0.9171 | 1.258 |
hu_maternal_loss4045 | 1.323 | 1.122 | 1.564 |
hu_older_siblings1 | 0.9877 | 0.8518 | 1.154 |
hu_older_siblings2 | 0.8528 | 0.6962 | 1.039 |
hu_older_siblings3 | 0.8198 | 0.6359 | 1.045 |
hu_older_siblings4 | 0.7922 | 0.5767 | 1.099 |
hu_older_siblings5P | 0.5601 | 0.3633 | 0.8548 |
hu_nr.siblings | 1.11 | 1.069 | 1.153 |
hu_last_born1 | 1.084 | 0.9586 | 1.221 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 2.27 | [1.91;2.65] | [2.02;2.51] |
estimate father 35y | 2.06 | [1.61;2.58] | [1.76;2.4] |
percentage change | -8.9 | [-26.92;11.29] | [-20.63;4.21] |
OR/IRR | 1.06 | [0.96;1.2] | [1;1.14] |
OR hurdle | 1.31 | [0.97;1.75] | [1.09;1.58] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r24_uniform_priors.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
In the three historical populations, records were kept in the parish. Although records were linked between parishes in all populations, except three out of four provinces in historical Sweden, migration might sometimes lead to censoring of records. Adjusting for migration may however constitute a partial adjustment for the outcome, as lower offspring fitness might make them more likely to migrate. Hence, we show the results of doing so as a robustness analysis. In all analyses, we adjusted for a “migrated”-dummy variable. Migration was differently defined depending on the population. In Québec, we had flags denoting immigrants and emigrants. Few immigrants were included in our analyses anyway, as we needed parental information for our analyses. Emigrants were people who left Québec. In historical Sweden, migration was logged as migration from the parish of birth. In the Krummhörn, we set migrated to true, when the parish of death/burial differed from the parish of birth/baptism.
No migration information was available in 20th-century Sweden, but records there weren’t kept in parishes, so this should not pose a problem.
model_summary = summary(model, use_cache = FALSE, priors = TRUE)
print(model_summary)
## Family: hurdle_poisson(log)
## Formula: children ~ paternalage + migrated + maternalage.factor + birth_cohort + male + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## hu ~ paternalage + migrated + maternalage.factor + birth_cohort + male + paternalage.mean + paternal_loss + maternal_loss + older_siblings + nr.siblings + last_born + (1 | idParents)
## Data: model_data (Number of observations: 9447)
## Samples: 6 chains, each with iter = 1000; warmup = 500; thin = 1;
## total post-warmup samples = 3000
## WAIC: Not computed
##
## Priors:
## b ~ normal(0,5)
## sd ~ student_t(3, 0, 5)
## b_hu ~ normal(0,5)
## sd_hu ~ student_t(3, 0, 10)
##
## Group-Level Effects:
## ~idParents (Number of levels: 2186)
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
## sd(Intercept) 0.23 0.01 0.20 0.25 1105 1.00
## sd(hu_Intercept) 0.70 0.06 0.59 0.81 821 1.01
##
## Population-Level Effects:
## Estimate Est.Error l-95% CI u-95% CI Eff.Sample
## Intercept 2.09 0.20 1.70 2.47 1745
## paternalage 0.07 0.05 -0.04 0.18 694
## migrated -0.44 0.18 -0.79 -0.08 1897
## maternalage.factor1420 -0.05 0.10 -0.23 0.14 3000
## maternalage.factor3550 0.00 0.03 -0.06 0.05 3000
## birth_cohort1760M1765 0.00 0.06 -0.12 0.12 1237
## birth_cohort1765M1770 -0.12 0.06 -0.23 0.00 963
## birth_cohort1770M1775 -0.11 0.06 -0.23 0.00 770
## birth_cohort1775M1780 -0.03 0.06 -0.13 0.08 727
## birth_cohort1780M1785 -0.11 0.06 -0.23 0.00 771
## birth_cohort1785M1790 -0.10 0.05 -0.20 0.01 646
## birth_cohort1790M1795 -0.08 0.05 -0.17 0.03 646
## birth_cohort1795M1800 -0.11 0.05 -0.21 -0.01 616
## birth_cohort1800M1805 -0.11 0.05 -0.21 -0.02 637
## birth_cohort1805M1810 -0.13 0.05 -0.23 -0.04 682
## birth_cohort1810M1815 -0.10 0.05 -0.19 0.00 601
## birth_cohort1815M1820 -0.14 0.05 -0.23 -0.05 554
## birth_cohort1820M1825 -0.18 0.05 -0.28 -0.09 546
## birth_cohort1825M1830 -0.21 0.05 -0.30 -0.12 539
## birth_cohort1830M1835 -0.18 0.05 -0.27 -0.09 563
## male1 0.08 0.02 0.04 0.11 3000
## paternalage.mean -0.07 0.06 -0.18 0.04 710
## paternal_loss01 -0.15 0.07 -0.30 -0.01 3000
## paternal_loss15 -0.04 0.05 -0.14 0.06 1733
## paternal_loss510 -0.07 0.04 -0.15 0.01 1757
## paternal_loss1015 0.01 0.04 -0.07 0.08 1597
## paternal_loss1520 -0.10 0.04 -0.17 -0.02 1727
## paternal_loss2025 -0.12 0.04 -0.20 -0.05 1562
## paternal_loss2530 -0.01 0.03 -0.08 0.05 1426
## paternal_loss3035 -0.03 0.03 -0.09 0.03 1529
## paternal_loss3540 -0.01 0.03 -0.07 0.05 1516
## paternal_loss4045 -0.01 0.04 -0.08 0.06 1968
## maternal_loss01 0.10 0.08 -0.05 0.25 3000
## maternal_loss15 -0.02 0.05 -0.11 0.07 2272
## maternal_loss510 0.07 0.04 -0.01 0.15 1912
## maternal_loss1015 0.03 0.04 -0.05 0.11 3000
## maternal_loss1520 0.00 0.04 -0.08 0.08 3000
## maternal_loss2025 0.01 0.04 -0.07 0.08 2258
## maternal_loss2530 -0.02 0.03 -0.09 0.05 2158
## maternal_loss3035 -0.05 0.03 -0.11 0.01 1987
## maternal_loss3540 -0.03 0.03 -0.09 0.03 2110
## maternal_loss4045 -0.03 0.03 -0.09 0.03 3000
## older_siblings1 0.03 0.03 -0.03 0.08 1427
## older_siblings2 -0.05 0.04 -0.12 0.02 847
## older_siblings3 -0.07 0.05 -0.16 0.02 794
## older_siblings4 -0.09 0.06 -0.20 0.03 819
## older_siblings5P -0.10 0.08 -0.25 0.06 697
## nr.siblings 0.01 0.01 0.00 0.02 944
## last_born1 -0.04 0.02 -0.09 0.00 3000
## hu_Intercept 5.80 0.47 4.90 6.73 1781
## hu_paternalage 0.11 0.18 -0.23 0.45 898
## hu_migrated -6.91 0.41 -7.77 -6.17 3000
## hu_maternalage.factor1420 0.12 0.29 -0.46 0.70 3000
## hu_maternalage.factor3550 0.03 0.09 -0.15 0.21 3000
## hu_birth_cohort1760M1765 -0.18 0.21 -0.60 0.23 3000
## hu_birth_cohort1765M1770 -0.34 0.19 -0.71 0.03 1210
## hu_birth_cohort1770M1775 0.03 0.18 -0.33 0.38 1056
## hu_birth_cohort1775M1780 -0.32 0.18 -0.68 0.02 1167
## hu_birth_cohort1780M1785 -0.53 0.19 -0.91 -0.16 1240
## hu_birth_cohort1785M1790 -0.38 0.18 -0.72 -0.03 1051
## hu_birth_cohort1790M1795 -0.34 0.17 -0.67 -0.01 918
## hu_birth_cohort1795M1800 -0.55 0.17 -0.88 -0.23 796
## hu_birth_cohort1800M1805 -0.69 0.16 -1.02 -0.38 844
## hu_birth_cohort1805M1810 -0.46 0.16 -0.79 -0.13 877
## hu_birth_cohort1810M1815 -0.36 0.15 -0.67 -0.07 709
## hu_birth_cohort1815M1820 -0.58 0.14 -0.87 -0.30 709
## hu_birth_cohort1820M1825 -0.43 0.15 -0.72 -0.13 755
## hu_birth_cohort1825M1830 -0.42 0.15 -0.69 -0.13 709
## hu_birth_cohort1830M1835 -0.37 0.15 -0.67 -0.08 680
## hu_male1 0.25 0.06 0.14 0.36 3000
## hu_paternalage.mean 0.00 0.18 -0.36 0.34 923
## hu_paternal_loss01 0.48 0.23 0.04 0.94 3000
## hu_paternal_loss15 0.49 0.16 0.18 0.81 1636
## hu_paternal_loss510 0.17 0.14 -0.10 0.45 1304
## hu_paternal_loss1015 0.18 0.14 -0.09 0.44 1385
## hu_paternal_loss1520 -0.07 0.13 -0.33 0.20 1386
## hu_paternal_loss2025 0.09 0.12 -0.16 0.33 1256
## hu_paternal_loss2530 0.02 0.12 -0.22 0.25 1271
## hu_paternal_loss3035 -0.07 0.11 -0.29 0.16 1210
## hu_paternal_loss3540 -0.01 0.11 -0.23 0.22 1439
## hu_paternal_loss4045 0.07 0.13 -0.18 0.32 3000
## hu_maternal_loss01 1.51 0.23 1.06 1.96 3000
## hu_maternal_loss15 0.65 0.15 0.34 0.95 3000
## hu_maternal_loss510 0.62 0.14 0.35 0.89 3000
## hu_maternal_loss1015 0.63 0.13 0.37 0.90 2098
## hu_maternal_loss1520 0.37 0.14 0.10 0.64 3000
## hu_maternal_loss2025 0.28 0.13 0.02 0.54 3000
## hu_maternal_loss2530 0.28 0.12 0.04 0.52 1741
## hu_maternal_loss3035 0.31 0.11 0.10 0.53 3000
## hu_maternal_loss3540 0.18 0.10 -0.02 0.37 2019
## hu_maternal_loss4045 0.30 0.11 0.09 0.53 3000
## hu_older_siblings1 0.00 0.09 -0.18 0.17 3000
## hu_older_siblings2 -0.16 0.12 -0.39 0.06 1088
## hu_older_siblings3 -0.13 0.15 -0.44 0.16 895
## hu_older_siblings4 0.00 0.19 -0.37 0.37 952
## hu_older_siblings5P -0.18 0.25 -0.67 0.30 883
## hu_nr.siblings 0.06 0.02 0.01 0.10 1051
## hu_last_born1 0.06 0.08 -0.10 0.21 3000
## Rhat
## Intercept 1.00
## paternalage 1.01
## migrated 1.00
## maternalage.factor1420 1.00
## maternalage.factor3550 1.00
## birth_cohort1760M1765 1.00
## birth_cohort1765M1770 1.00
## birth_cohort1770M1775 1.00
## birth_cohort1775M1780 1.00
## birth_cohort1780M1785 1.00
## birth_cohort1785M1790 1.01
## birth_cohort1790M1795 1.00
## birth_cohort1795M1800 1.00
## birth_cohort1800M1805 1.00
## birth_cohort1805M1810 1.00
## birth_cohort1810M1815 1.00
## birth_cohort1815M1820 1.00
## birth_cohort1820M1825 1.00
## birth_cohort1825M1830 1.00
## birth_cohort1830M1835 1.00
## male1 1.00
## paternalage.mean 1.01
## paternal_loss01 1.00
## paternal_loss15 1.00
## paternal_loss510 1.00
## paternal_loss1015 1.00
## paternal_loss1520 1.00
## paternal_loss2025 1.00
## paternal_loss2530 1.00
## paternal_loss3035 1.00
## paternal_loss3540 1.00
## paternal_loss4045 1.00
## maternal_loss01 1.00
## maternal_loss15 1.00
## maternal_loss510 1.00
## maternal_loss1015 1.00
## maternal_loss1520 1.00
## maternal_loss2025 1.00
## maternal_loss2530 1.00
## maternal_loss3035 1.00
## maternal_loss3540 1.00
## maternal_loss4045 1.00
## older_siblings1 1.00
## older_siblings2 1.01
## older_siblings3 1.01
## older_siblings4 1.01
## older_siblings5P 1.01
## nr.siblings 1.00
## last_born1 1.00
## hu_Intercept 1.00
## hu_paternalage 1.00
## hu_migrated 1.00
## hu_maternalage.factor1420 1.00
## hu_maternalage.factor3550 1.00
## hu_birth_cohort1760M1765 1.00
## hu_birth_cohort1765M1770 1.00
## hu_birth_cohort1770M1775 1.00
## hu_birth_cohort1775M1780 1.00
## hu_birth_cohort1780M1785 1.00
## hu_birth_cohort1785M1790 1.00
## hu_birth_cohort1790M1795 1.00
## hu_birth_cohort1795M1800 1.00
## hu_birth_cohort1800M1805 1.00
## hu_birth_cohort1805M1810 1.00
## hu_birth_cohort1810M1815 1.00
## hu_birth_cohort1815M1820 1.00
## hu_birth_cohort1820M1825 1.01
## hu_birth_cohort1825M1830 1.00
## hu_birth_cohort1830M1835 1.01
## hu_male1 1.00
## hu_paternalage.mean 1.00
## hu_paternal_loss01 1.00
## hu_paternal_loss15 1.00
## hu_paternal_loss510 1.00
## hu_paternal_loss1015 1.00
## hu_paternal_loss1520 1.00
## hu_paternal_loss2025 1.00
## hu_paternal_loss2530 1.00
## hu_paternal_loss3035 1.00
## hu_paternal_loss3540 1.00
## hu_paternal_loss4045 1.00
## hu_maternal_loss01 1.00
## hu_maternal_loss15 1.00
## hu_maternal_loss510 1.00
## hu_maternal_loss1015 1.00
## hu_maternal_loss1520 1.00
## hu_maternal_loss2025 1.00
## hu_maternal_loss2530 1.00
## hu_maternal_loss3035 1.00
## hu_maternal_loss3540 1.00
## hu_maternal_loss4045 1.00
## hu_older_siblings1 1.00
## hu_older_siblings2 1.00
## hu_older_siblings3 1.00
## hu_older_siblings4 1.00
## hu_older_siblings5P 1.00
## hu_nr.siblings 1.00
## hu_last_born1 1.00
##
## Samples were drawn using sampling(NUTS). For each parameter, Eff.Sample
## is a crude measure of effective sample size, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).
Estimates are exp(b)
. When they are referring to the hurdle (hu) component, or a dichotomous outcome, they are odds ratios, when they are referring to a Poisson component, they are hazard ratios. In both cases, they are presented with 95% credibility intervals. To see the effects on the response scale (probability or number of children), consult the marginal effect plots.
fixed_eff = data.frame(model_summary$fixed, check.names = F)
fixed_eff$Est.Error = fixed_eff$Eff.Sample = fixed_eff$Rhat = NULL
fixed_eff$`Odds/hazard ratio` = exp(fixed_eff$Estimate)
fixed_eff$`OR/HR low 95%` = exp(fixed_eff$`l-95% CI`)
fixed_eff$`OR/HR high 95%` = exp(fixed_eff$`u-95% CI`)
fixed_eff = fixed_eff %>% select(`Odds/hazard ratio`, `OR/HR low 95%`, `OR/HR high 95%`)
pander::pander(fixed_eff)
Odds/hazard ratio | OR/HR low 95% | OR/HR high 95% | |
---|---|---|---|
Intercept | 8.07 | 5.459 | 11.77 |
paternalage | 1.072 | 0.9635 | 1.192 |
migrated | 0.6462 | 0.454 | 0.9233 |
maternalage.factor1420 | 0.9507 | 0.7942 | 1.149 |
maternalage.factor3550 | 0.997 | 0.9452 | 1.054 |
birth_cohort1760M1765 | 0.9979 | 0.8831 | 1.126 |
birth_cohort1765M1770 | 0.8874 | 0.7922 | 0.9958 |
birth_cohort1770M1775 | 0.8922 | 0.7958 | 1.004 |
birth_cohort1775M1780 | 0.9709 | 0.8747 | 1.085 |
birth_cohort1780M1785 | 0.897 | 0.7981 | 1.003 |
birth_cohort1785M1790 | 0.9076 | 0.8174 | 1.007 |
birth_cohort1790M1795 | 0.9275 | 0.8405 | 1.028 |
birth_cohort1795M1800 | 0.8967 | 0.8136 | 0.9899 |
birth_cohort1800M1805 | 0.8923 | 0.808 | 0.983 |
birth_cohort1805M1810 | 0.8738 | 0.7926 | 0.9636 |
birth_cohort1810M1815 | 0.9086 | 0.8295 | 0.9969 |
birth_cohort1815M1820 | 0.868 | 0.7934 | 0.9525 |
birth_cohort1820M1825 | 0.8317 | 0.759 | 0.9114 |
birth_cohort1825M1830 | 0.811 | 0.738 | 0.888 |
birth_cohort1830M1835 | 0.8333 | 0.7598 | 0.9163 |
male1 | 1.081 | 1.046 | 1.116 |
paternalage.mean | 0.9318 | 0.8333 | 1.04 |
paternal_loss01 | 0.8576 | 0.7428 | 0.989 |
paternal_loss15 | 0.9606 | 0.87 | 1.058 |
paternal_loss510 | 0.9328 | 0.8603 | 1.009 |
paternal_loss1015 | 1.006 | 0.9297 | 1.087 |
paternal_loss1520 | 0.9069 | 0.8422 | 0.9761 |
paternal_loss2025 | 0.8854 | 0.8216 | 0.9508 |
paternal_loss2530 | 0.987 | 0.9237 | 1.053 |
paternal_loss3035 | 0.9708 | 0.9095 | 1.035 |
paternal_loss3540 | 0.9863 | 0.9286 | 1.051 |
paternal_loss4045 | 0.9891 | 0.9212 | 1.06 |
maternal_loss01 | 1.109 | 0.9496 | 1.283 |
maternal_loss15 | 0.9828 | 0.8955 | 1.075 |
maternal_loss510 | 1.074 | 0.9908 | 1.159 |
maternal_loss1015 | 1.029 | 0.9495 | 1.116 |
maternal_loss1520 | 1.003 | 0.9275 | 1.083 |
maternal_loss2025 | 1.007 | 0.9309 | 1.086 |
maternal_loss2530 | 0.9787 | 0.9152 | 1.05 |
maternal_loss3035 | 0.9525 | 0.8918 | 1.014 |
maternal_loss3540 | 0.9681 | 0.9135 | 1.027 |
maternal_loss4045 | 0.9716 | 0.9143 | 1.032 |
older_siblings1 | 1.026 | 0.9739 | 1.083 |
older_siblings2 | 0.9525 | 0.888 | 1.023 |
older_siblings3 | 0.9291 | 0.8491 | 1.022 |
older_siblings4 | 0.9118 | 0.8154 | 1.028 |
older_siblings5P | 0.9068 | 0.7804 | 1.059 |
nr.siblings | 1.01 | 0.9956 | 1.024 |
last_born1 | 0.9575 | 0.9145 | 1.002 |
hu_Intercept | 330 | 134.9 | 835.2 |
hu_paternalage | 1.115 | 0.7967 | 1.572 |
hu_migrated | 0.0009989 | 0.0004209 | 0.002101 |
hu_maternalage.factor1420 | 1.133 | 0.6328 | 2.004 |
hu_maternalage.factor3550 | 1.031 | 0.8579 | 1.234 |
hu_birth_cohort1760M1765 | 0.8369 | 0.5512 | 1.26 |
hu_birth_cohort1765M1770 | 0.7126 | 0.4899 | 1.03 |
hu_birth_cohort1770M1775 | 1.033 | 0.7218 | 1.464 |
hu_birth_cohort1775M1780 | 0.7227 | 0.5083 | 1.025 |
hu_birth_cohort1780M1785 | 0.5902 | 0.4025 | 0.8543 |
hu_birth_cohort1785M1790 | 0.6825 | 0.4845 | 0.9674 |
hu_birth_cohort1790M1795 | 0.711 | 0.5096 | 0.9909 |
hu_birth_cohort1795M1800 | 0.5755 | 0.4154 | 0.7944 |
hu_birth_cohort1800M1805 | 0.5012 | 0.362 | 0.685 |
hu_birth_cohort1805M1810 | 0.6337 | 0.4553 | 0.8744 |
hu_birth_cohort1810M1815 | 0.6961 | 0.514 | 0.9357 |
hu_birth_cohort1815M1820 | 0.5609 | 0.4209 | 0.7376 |
hu_birth_cohort1820M1825 | 0.6509 | 0.4863 | 0.8807 |
hu_birth_cohort1825M1830 | 0.6602 | 0.5009 | 0.8771 |
hu_birth_cohort1830M1835 | 0.69 | 0.5095 | 0.9258 |
hu_male1 | 1.286 | 1.15 | 1.439 |
hu_paternalage.mean | 0.9981 | 0.7008 | 1.4 |
hu_paternal_loss01 | 1.624 | 1.04 | 2.562 |
hu_paternal_loss15 | 1.63 | 1.198 | 2.251 |
hu_paternal_loss510 | 1.189 | 0.9054 | 1.562 |
hu_paternal_loss1015 | 1.192 | 0.9143 | 1.557 |
hu_paternal_loss1520 | 0.934 | 0.7213 | 1.219 |
hu_paternal_loss2025 | 1.093 | 0.856 | 1.391 |
hu_paternal_loss2530 | 1.016 | 0.8054 | 1.282 |
hu_paternal_loss3035 | 0.9325 | 0.7466 | 1.17 |
hu_paternal_loss3540 | 0.9937 | 0.7955 | 1.244 |
hu_paternal_loss4045 | 1.068 | 0.836 | 1.377 |
hu_maternal_loss01 | 4.543 | 2.9 | 7.103 |
hu_maternal_loss15 | 1.916 | 1.412 | 2.587 |
hu_maternal_loss510 | 1.863 | 1.425 | 2.431 |
hu_maternal_loss1015 | 1.884 | 1.45 | 2.459 |
hu_maternal_loss1520 | 1.445 | 1.101 | 1.891 |
hu_maternal_loss2025 | 1.322 | 1.021 | 1.708 |
hu_maternal_loss2530 | 1.32 | 1.044 | 1.676 |
hu_maternal_loss3035 | 1.368 | 1.103 | 1.701 |
hu_maternal_loss3540 | 1.191 | 0.9766 | 1.449 |
hu_maternal_loss4045 | 1.351 | 1.095 | 1.692 |
hu_older_siblings1 | 1 | 0.8372 | 1.19 |
hu_older_siblings2 | 0.8505 | 0.6764 | 1.067 |
hu_older_siblings3 | 0.8746 | 0.6465 | 1.176 |
hu_older_siblings4 | 1.003 | 0.6875 | 1.442 |
hu_older_siblings5P | 0.8385 | 0.5127 | 1.353 |
hu_nr.siblings | 1.06 | 1.014 | 1.106 |
hu_last_born1 | 1.059 | 0.9084 | 1.239 |
pander::pander(paternal_age_10y_effect(model))
effect | median_estimate | ci_95 | ci_80 |
---|---|---|---|
estimate father 25y | 0.93 | [0.63;1.32] | [0.72;1.17] |
estimate father 35y | 0.9 | [0.58;1.35] | [0.68;1.19] |
percentage change | -1.98 | [-28.8;31.32] | [-20.51;19.12] |
OR/IRR | 1.07 | [0.96;1.19] | [1;1.15] |
OR hurdle | 1.11 | [0.8;1.57] | [0.9;1.4] |
In these marginal effect plots, we set all predictors except the one shown on the X axis to their mean and in the case of factors to their reference level. We then plot the estimated association between the X axis predictor and the outcome on the response scale (e.g. probability of survival/marriage or number of children).
plot.brmsMarginalEffects_shades(
x = marginal_effects(model, re_formula = NA, probs = c(0.025,0.975)),
y = marginal_effects(model, re_formula = NA, probs = c(0.1,0.9)),
ask = FALSE)
Here, we plotted the 95% posterior densities for the unexponentiated model coefficients (b_
). The darkly shaded area represents the 50% credibility interval, the dark line represent the posterior mean estimate.
mcmc_areas(as.matrix(model$fit), regex_pars = "b_[^I]", point_est = "mean", prob = 0.50, prob_outer = 0.95) + ggtitle("Posterior densities with means and 50% intervals") + analysis_theme + theme(axis.text = element_text(size = 12), panel.grid = element_blank()) + xlab("Coefficient size")
These plots were made to diagnose misfit and nonconvergence.
In posterior predictive checks, we test whether we can approximately reproduce the real data distribution from our model.
brms::pp_check(model, re_formula = NA, type = "dens_overlay")
brms::pp_check(model, re_formula = NA, type = "hist")
Did the 6 chains converge?
stanplot(model, pars = "^b_[^I]", type = 'rhat')
stanplot(model, pars = "^b", type = 'neff_hist')
Trace plots are only shown in the case of nonconvergence.
if(any( summary(model)$fixed[,"Rhat"] > 1.1)) { # only do traceplots if not converged
plot(model, N = 3, ask = FALSE)
}
This model was stored in the file: coefs/krmh/r25_migration_status.rds.
Click the following link to see the script used to generate this model:
opts_chunk$set(echo = FALSE)
clusterscript = str_replace(basename(model_filename), "\\.rds",".html")
cat("[Cluster script](" , clusterscript, ")", sep = "")
In this model, we adjust for maternal age using a continuous variable. We also adjust for a dummy variable for teenage motherhood, to account for the nonlinearity of the maternal age effect. Moreover, we use separate random intercepts for mothers and fathers and adjust for the mother’s mean age at birth and the father’s mean age at birth. This model only converges in the 20th-century Sweden data, because there are sufficient numbers of divorces and remarriages and enough data to separate the parents’ contributions.
Here we show the effect of paternal age for each episode.
In reference to m3, the main reported model, the robustness models were implemented as follows: r1 relaxed exclusion criteria (not in 20th-century Sweden), r2 had only birth cohort as a covariate, r3 adjusted for birth order as a continuous variable, r4 adjusted for number of dependent siblings instead of birth order, r5 interacted birth order with number of siblings, r6 did not adjust for birth order, r7 adjusted only for parental loss in the first 5 years, r8 adjusted for being the first-/last-born adult son, r9 adjusted for a continuous nonlinear thin-splate spline for birth year instead of 5-year bins, r10 added a group-level slope for paternal age, r11 included separate group-level effects for each parent instead of one per marriage, r12 added a moderation by anchor sex, r13 adjusted for paternal age at first birth, r14 compared a model with linear group fixed effects, r15 added a moderator by region and group-level effects by church parish (not in 20th-century Sweden), r16 was restricted to Skellefteå (only in historical Sweden), r17 simulated Down syndrome cases, r18 reversed hurdle Poisson and Poisson distribution for the respective populations, r19 used a normal distribution, r20 did not adjust for maternal age, r21 adjusted for maternal age as a continuous variable, r22 relaxed exclusion criteria and included 30 more years of birth cohorts, allowing for more potential censoring, r23 used Student’s t distributions for population-level priors and half-Cauchy priors for the family variance component, r24 used noninformative priors, which should lead to results comparable with maximum likelihood, r25 controlled for migration status (not in 20th-century Sweden), r26 separate parental age contributions (only in 20th-century Sweden).
Points show median estimates, the lines show 80% and 95% credibility intervals respectively.