To evaluate differences in the model across genotypes, three steps were undertaken. First, to obtain baseline information, a saturated model was conducted on the entire sample (e.g., collapsed across genotype). Second, because our moderator variable was dichotomous, a multiple group structural equation model was conducted using DRD4 genotype (7− vs. 7+) as the grouping variable (see Cortina, Chen, & Dunlap, 2001). Paths that were nonsignificant in both groups were trimmed from the model. Because hypotheses stemming from DST intrinsically involve moderation, paths that were significant in only one or both groups were retained. Thus, in the third step the final trimmed model was analyzed. Chi-square difference tests were used to determine group differences. Moderated paths were further tested for reliability through bootstrapping standard errors (n = 1,000 draws) and constructing 95% confidence intervals (CIs). Bootstrapped CIs estimate the population distribution of possible unstandardized parameter estimates based on n random draws from the analysis sample (see Mooney & Duval, 1993). Parameter reliability is evidenced when the bootstrapped 95% CI does not include zero. Last, bootstrapping was similarly implemented to test indirect effects (see MacKinnon, Lockwood, & Williams, 2004).
