The present sample allows several methods to estimate heritability of the phenotypes. First, a large portion of the sample is composed of twins. We used knowledge of twin zygosity and standard biometric statistical models to compute the additive genetic (A), shared environmental (C), and non-shared environmental (E) components of the 5x5 variance-covariance matrix of our five measures. This is the standard multivariate ACE model (Neale & Cardon, 1992). Variance-covariance component matrices were Cholesky-factorized and estimated by full information maximum likelihood after correcting for fixed effects of sex, age, year of birth, generational status, and the first 10 genetic principle components computed from EIGENSTRAT. Model fit was evaluated with accepted indexes of fit, including a likelihood ratio test and the Akaike Information Criterion (AIC). The likelihood ratio test is sensitive to sample size and correlational magnitude, and so is often augmented with measures of fit like the AIC, which have attractive theoretical properties not shared by the likelihood ratio, such as minimization of mean squared error of estimation (Vrieze, 2012).
