Heavy episodic drinking data were first grouped by age, resulting in 14 separate points of measurement from age 15 to 28. However, given the wide age range in the sample and the fairly restricted sample size, this approach resulted in a large amount of missing data for certain ages. To partially alleviate this concern, and given that assessments were conducted biennially, data were combined into two-year intervals to create 7 separate points of measurement (ages 15.5, 17.5, 19.5, 21.5, 23.5, 25.5 and 27.5). A generalized linear mixed model with log link and Poisson distribution was used to model the trajectory of heavy episodic drinking (Figure 2). Linear, quadratic, and linear piecewise growth parameters were estimated and compared, in order to determine an optimal model based on information criteria fit statistics (i.e., AIC and BIC). The piecewise model was specified as followed: the first portion of the slope reflected the change in heavy episodic drinking from adolescence (age 15.5) to early adulthood (age 21.5), and the second portion of the slope reflects adult trajectories of heavy episodic drinking, from age 23.5
