To utilize fully the information in our cohort, and to accommodate covariate effects, we estimated heritabilities using a variance components model [23,24]. This model can accommodate pedigrees of any configuration and is well suited to the analysis of extended pedigrees. We included age, age2, sex, and the two corresponding interaction terms as covariates in all analyses. We considered a series of models for each trait, and these are detailed below, together with details of transformations we applied to each trait to guard against the possibility of statistical artifacts induced by outliers or non-normal marginal distributions.
