Analyses were conducted in SAS v9.3. To reduce the burden of multiple measures testing, a principal components analysis was used to combine placebo-adjusted indices for the 7 subjective measures into a smaller number of factors. Principal component, rather than exploratory factor analysis, was employed as the recommended approach for data reduction and in order to maximize the total variance accounted for, as no underlying latent construct was hypothesized for this variable set (Fabrigar, 1999; Brown, 2000). Principal components accounting for at least 20% of the variability in the aggregate measure were included as potential phenotypes of interest. Potential confounds due to differences in demographic variables were assessed. Chi-squared tests assessed differences between pairs of categorical demographic variables (family history of alcoholism, gender, session order). Mantel-Haenszel tests were used to test if categorical demographic variables differed by SNP genotype. T-tests were employed to test if drinks per drinking day and age differed by gender or family history of alcoholism. Analysis of variance tested for the additive effects of each SNP genotype on drinks per drinking day and age.
