In each of the 6 cohorts with twin data separately, twin correlations for the IRT latent trait scores were estimated using the structural equation modeling package OpenMx within the statistical software program R (Boker et al. 2011). This was done by fitting a fully saturated model using full information likelihood to the data of twins in five sex-by-zygosity groups: monozygotic male twin pairs (MZM), dizygotic male twin pairs (DZM), monozygotic female twin pairs (MZM), dizygotic female twin pairs (DZM) and dizygotic twin pairs of opposite sex (DOS; if available in the particular cohort). Twin pairs in which Neuroticism and Extraversion scores were available for both twins were included, as well as twin pairs for which information was available for only one of the twins. In each cohort including a DOS group, 16 parameters were estimated: 5 means (5 sex by zygosity groups), 1 regression parameter for the effect of age on the means, 5 variances (5 sex by zygosity groups) and 5 covariances (for 5 sex by zygosity groups). In the cohorts without a DOS group, 4 means, 1 regression
