the “average number of drinks” might be more reflective of overall drinking patterns for an individual. In this sample, the “max drinks” phenotype is modestly correlated with “average drinks” (r2 = 0.25; p < 0.001). In an effort to examine why our results might have differed from those previously reported a post-hoc examination of “max drinks” was undertaken. As expected, there was no evidence for association with any of the SNPs (all p>0.065). These results underscore the importance of phenotype definition and illustrate the possible limitations of approaches focused on clinical diagnosis or simple quantity/frequency measures. They also illustrate the utility of using a family-based approach with elegant statistical tools such as those implemented in FBAT-PC to reduce the dimensionality of the data and obtain informative associations with composite phenotypes.
