Some study limitations are noteworthy. First, young adult participants were asked to recall their parenting and peer influences retrospectively, whereas adolescents reported on these effects concurrently. Although participants were followed longitudinally, they did not contribute to data across the entire age range in the study (age 15 to 28). To characterize the heterogeneity of heavy episodic drinking, a growth mixture model would have been preferred. Limited data precluded this analysis, although data collection in the Prospective Study is on-going and we anticipate on having enough data with which to perform more rigorous examinations of growth mixture models that extend beyond age 28 in future investigations. Second, the participants in the Prospective Study were genetically related to the participants of the original COGA study, which was a case-control designed study (proband status of the relatives was not controlled for in the current analyses). Thus, the derivation of PRS did not involve a completely independent discovery and target dataset. As an overlap between these datasets may potentially lead to an overestimation of the prediction accuracy for the PRS (Wray et al., 2013),
