The twin model we employed has been described in detail elsewhere (Dick et al., 2011). Briefly, a multivariate Cholesky model was fit to the measures of alcohol consumption and problems in order to estimate (1) the magnitude of genetic and environmental influences on each phenotype and (2) the extent to which these influences contributed to the covariation between the phenotypes. Using the statistical software package Mx (Neale and Cardon, 1992), we generated individual scores for each subject weighted by the loadings implicated by the genetic architecture from the best fitting twin model. When the best fitting model (Figure 1) from the full sample (n=2,500) was fit in the genotyped subset (n=602), there was not a significant decrease in model fit (χ2=3.28, p=1.00), however a model with two genetic factors (A1 and A6) fit the genotyped subsample best (AIC = −352.334). Thus, we moved two genetic factors forward in creating individual genetic factor scores for each person within the genotyped sample; (1) A general factor which loads onto measures of alcohol consumption and problems and (2) an alcohol problems factor which
