Indirect effects of alc-GPS on alcohol use via sensation seeking, family and friend support were tested using the MODEL INDIRECT command in Mplus, which provides a test of specific indirect effects in addition to the total indirect and direct effects of alc-GPS on alcohol use. Indirect effects were evaluated using bias-corrected bootstrapping (1,000 times) 95% confidence intervals (CI; MacKinnon et al., 2004), with CI not including zero indicating statistically significant indirect effects. Clustering within families was accounted for using the CLUSTER command in Mplus. Maximum likelihood with robust standard errors (MLR) estimation was used in all analysis to account for non-normality and non-independence of the observed data. Missing data was accounted for using full information maximum likelihood.
