Third, to account for the possibility of gene-environment correlation, which may confound the interpretation of potential GxE effects (Knafo and Jaffee, 2013) we examined correlations between PRS and each of the environmental variables assessed in the study. PRS was significantly associated with close friend substance use (r =.23, p < .01), parental knowledge (r =−.18, p < .01) and school substance use (r =.16, p < .01). To address concerns that the GxE results were impacted by the presence of G-E correlation, we used standardized residual values from a regression model where each environmental variable was separately regressed on PRS. The use of residualized variables statistically eliminates G-E correlation effects in the model because the genetic and environmental effects have been partialed from one another (Salvatore et al., 2014b). Parallel analyses (i.e., hierarchical poisson regressions) were conducted in which growth parameters (i.e., intercepts and slopes) were regressed on to the standardized residual values for close friend substance use, PRS and their interaction, controlling for the standardized residual values for parental knowledge and school substance use. Results from this analysis was
