Next, we conducted path analysis (see Figure 1 for the conceptual model). Externalizing PGS was specified to have a direct genetic effect on T1 externalizing (as indicated by path a in Figure 1), a direct genetic effect on T1 parental knowledge (path b), and a direct genetic effect on T1 peer substance use (path c). Paths b and c test for evidence of gene-environment correlations. In addition, adolescent externalizing behavior, parental knowledge, and peer substance use were specified to be correlated within T1. Externalizing PGS was specified to predict T2 externalizing behavior directly (not shown), and indirectly through T1 parental knowledge (path d) and T1 peer substance use (path e), while controlling for previous levels of externalizing behavior at T1. Additional cross-domain paths between T1 and T2 were also included: associations between T1 externalizing behavior and T2 parental knowledge (path f) and T2 peer substance use (path g); association between T1 parental knowledge and T2 peer substance use (path h); and association between T1 peer substance use and T2 parental knowledge (path i).
