Next, we examined an alternative model where social support moderates the associations between alc-GPS and alcohol use, as well as the association between sensation seeking and alcohol use (Figure 1, Panel B). In this model, alc-GPS was specified to be associated with alcohol use directly and indirectly via sensation seeking. Sensation seeking, family and friend support were specified to be associated with alcohol use. In addition, product terms between alc-GPS and family/friend support, as well as product terms between sensation seeking and family/friend support were included as additional independent variables associated with alcohol use; variables were mean-centered before creating the product terms. alc-GPS and family/friend support were specified to be correlated using the “WITH” command in Mplus to account for potential rGE. Follow-up analysis were conducted to check the robustness of significant GxE effects by including covariates ~ G and covariates ~ E interaction terms in the model that further account for potential confounding effects (Keller, 2012).
