The fit of the model with heavy drinking (Figure 3) also was acceptable [χ2(69)=l 57.26, p<.00l; CFI=.89, RMSEA=.06, SRMR=.06]. With the exception of friends' alcohol use between adolescence and emerging adulthood, there was significant continuity in both heavy drinking and friends' alcohol use across adjacent time points; continuity was stronger during adulthood compared to the adolescence to adulthood transition. Friends' alcohol use and heavy drinking were positively correlated within each wave. Heavy drinking predicted higher proportion of heavy drinking friends at the next time point across all ages, indicating developmentally consistent friend selection effects. In addition, friends' alcohol use at age 17 predicted more frequent heavy drinking at age 23, suggesting a peer socialization effect developmentally specific to the transition from adolescence to emerging adulthood (see Figure 3). Multigroup modeling indicated significant model differences between the low and high risk genotype groups [Δχ2(16)=28.99, p<.05]. However, none of the follow up tests reached significance level after the Bonferroni correction was applied.
