The next series of conditional GMM was run incorporating both genetic and environmental covariates, wherein probability of class membership was regressed on (1) genotype, (2) peer group antisocial behavior, and (3) a product term representing the genotype by environment interaction. Again, multinomial logistic regression analyses were used to determine whether genotype and peer group antisocial behavior functioned interdependently in relation to probabilistic patterns of development in externalizing behavior. Table 5 provides evidence of significant moderation. Findings indicated that the overall level of antisocial behavior within the extended peer group moderated the extent to which genotype was predictive of trajectory class membership. Specifically, peer group antisocial behavior was differentially associated with CHRM2 genotype in adolescents with stable high externalizing trajectories when compared to those with more normatively decreasing low levels of problem behavior. This effect was demonstrated in relation to six of the nine candidate SNPs (Table 5; odds ratios ranging from 2.6 to 3.6), with a trend toward significance in a seventh SNP (OR = 2.2). In addition, peer group antisocial behavior was useful in distinguishing between genotypic variability in
