Independent t-tests and Chi-square tests were used to compare cannabis-dependent and control subjects according to sociodemographic characteristics. Association between genotype, behavioral traits and cannabis-use outcomes were analyzed by ANOVA and Pearson correlation, while group (based on cannabis-dependence diagnosis) and gene x group interaction effects were calculated using a general linear model. For each SNP, polymorphisms that were significantly associated with phenotype were evaluated for genotyping patterns of variants using the Fisher’s exact test. A reduction in unstandardized beta of >10% on cannabis dependence by the genetic factor after the potential mediator (e.g., Neuroticism) was included in the model, and was considered to be appreciably different and reported as a partial mediator. Additive interaction was evaluated based on the Rothman “index of synergism.” The presence/absence of an additive interaction was examined using an index, attributable proportion due to interaction (AP) and the 95% confidence interval (CI) estimated based on the Hosmer-Lemeshow CI estimation of interaction. AP exceeding zero indicates increased risk due to the two risk factors. Thus, the 95% CI for an AP that does not include a value of
