To determine the stability of the association between drinking status and risky behaviors after adjusting for relevant demographic characteristics, multivariable GEE models were explored. An a priori decision was made that race and family type should be included in all multivariable models to account for the family ascertainment strategy. Additional potential covariates (i.e., gender, income, and age) were entered individually into a GEE model that already included race and family type. Additional covariates that were significant in these models were included together in a multivariable model. Chi-square statistics are reported for each predictor variable included in the GEE model, after adjusting for all variables in the model. When there was a significant group effect and within the context of the overall GEE model, statistical contrasts were used to assess pairwise comparisons between drinking groups. Adjusted odds ratios (ORs) with 95% confidence intervals are reported for the multivariable models. ORs reflect the increased odds of the presence of the risky behavior as compared to the lower ordered drinking group, after adjusting for all other variables in the model. Nondrinkers are used
