We conducted analyses in 3 stages. We first determined the longitudinal course of heavy episodic drinking over the course of time, comparing the goodness of fit to linear and polynomial models (quadratic). We used model fit statistics, including the Akaike information criterion (AIC) and likelihood ratio tests to assess improvement in fit. Second, we examined the appropriate structure of the random-effects components of the model using nested models. We estimated random effects using restricted maximum likelihood (REML). We used AIC and likelihood ratio tests to assess model fit improvement. Finally, we examined our main research questions (focal analyses)—that is, do PGSPAU, parental separation and parental relationship discord influence the longitudinal course of heavy episodic drinking from adolescence to young adulthood? Do parental separation and parental relationship discord modify the influence of PGSPAU on the longitudinal course of heavy episodic drinking over time? Prior to running the focal analyses, all continuous predictors were centered at their means, which improves the interpretability of the intercept. Furthermore, we examined the interactive effects between PGSPAU and parental separation and parental relationship discord by including
