The results of the lagged logistic regression for girls are presented in Table 2. Model 1 showed strong growth in alcohol use by girls from waves 3–5 (reported above). In Model 2, lagged family variables were added. This resulted in a reduction in the age effect of about 20%. The effect for emotional closeness to mother was significant (OR = .69, p < .05). The effects for parent disapproval of alcohol use, emotional closeness to fathers, and family conflict were nonsignificant in Model 2. With peer alcohol use in the model (Model 3), parent disapproval and family conflict became significant (p < .05). These effects were weak (dependent on other variables), suggesting a potential suppressor effect for peer alcohol use or a highly correlated other variable. These effects were unlikely to be due to multicollinearity (mean Variance Inflation Factor = 1.28; [35]). Peer alcohol use strongly predicted the frequency of alcohol use, with ORs increasing nonlinearly with the number of peers who consumed alcohol (Table 2/3). To verify the appropriate conditions for mediation were present for the association of parental
