We ran the analyses in the nlme package version 3.1–117 (Pinheiro et al., 2014) in R version 3.0.3 (R Development Core Team, 2014). In order to account for autoregressive effects in our models, we specified a “corCAR1” correlation structure (i.e., an autocorrelation structure of order 1, with a continuous time covariate). Preliminary analyses examining the shape of the CBCL externalizing trajectory across time indicated a decreasing linear trajectory; accordingly, a linear time term was included in the model. A linear time term was also included in the TRF externalizing models in view of preliminary analyses that indicated an increasing level of externalizing behavior between ages 12–13. Time was centered at age 12 in all models to aid in the interpretation of the intercepts. GABRA2 genotype was coded 0, 1, or 2, representing the number of copies of the minor allele. The life events, parental antisocial behavior, and parental alcohol problems independent variables were standardized. Effect size was calculated as the percent of the variance (R2) accounted for by a parameter of interest above and beyond the covariates and main effects (for interactions) using the residual variance procedure described by Selya et al. (2012).
