Repeated measures latent class analysis (RMLCA)[38] was used to identify seven classes (i.e., patterns) of drinking across 12 weeks of treatment, as described elsewhere[6]. RMLCA is a latent variable mixture model in which the indicators of the latent class are repeated measures (e.g., weekly drinking). The current study examined demographic characteristics (age, sex (male=1), marital status (married=1), race/ethnicity (non-Hispanic White=1)), baseline percentage of heavy drinkers in the social network, drinks per day in the week prior to the baseline assessment, baseline alcohol dependence severity, and baseline negative mood symptoms, as well as all possible two-way and three-way interactions, as predictors of the seven patterns of drinking during treatment. We initially tested all possible two-way and three-way interactions using an a priori criterion for interaction effects of p<.01. However, no three-way interactions were retained and only two of the two-way interactions were retained: (1) age-by-baseline drinking, and (2) age-by-mood symptoms. Significant two-way interactions were probed using simple slopes analysis[39]. In addition, the effect of study membership and two-way interactions between study membership and covariates were also included in all models (Supplementary
