Based on the results of fitting linear regression models, we subsequently modified the multivariate variance components model to control for age as well as significant prenatal and parental predictors (p < .05). This was done by jointly modeling the linear regression of outcome (i.e., INATT, HYP/IMP, CDP, or AlcProb) on these covariates and the genetic and environmental contributions to the residual variance and covariance among outcome symptom count scores. In order to control for the age range in these data, we modeled age as a contrast coded covariate allowing for three groups: 11-14 years old, 15-18 years old, and 19+ years old. Models were fitted by maximum-likelihood using Mx (Neale et al., 1992). Under this adjusted means model, genetic (additive and dominant) and environmental (shared and nonshared) parameter estimates were obtained after controlling for significant predictors of each outcome. By doing this, we tested for residual genetic and environmental contributions to variation in risk of INATT, HYP/IMP, CDP, and AlcProb, as well as residual genetic and environmental correlations among our four phenotypes. These models allowed for a contrast effect for ADHD subtype scores (HYP/IMP and INATT).
