To examine the relationship between AA attendance, the four network variables (i.e., pro-abstinence, pro-drinking, abstinent activities, drinking activities), and alcohol use, we used a general linear model (GLM) controlling for baseline covariates. We ran these models separately for the aftercare and outpatient samples and for each alcohol use outcome variable (PDA, DDD) to examine: (1) the independent (AA attendance) to dependent variable path, (2) the independent to mediator (network support/activities) path, and (3) the mediator to dependent variable path. To make the tests prospective (lagged), we used time points in which predictors were used for outcomes (i.e., AA attendance past 90 days at month 3 to predict network support/activities at month 9 [past 6 months] and alcohol use at month 15 [past 90 days]; similarly, we used network support/activities at month 9 to predict drinking at month 15 [past 90 days]). The static covariates were age, ethnicity, gender, marital status, employment status, number of prior alcohol-related treatments, treatment assignment, treatment site, the relevant baseline level of the dependent variable (i.e., network support/activities, PDA, DDD). Figure 1 shows this model.
