For all analyses, we were testing specific hypotheses, and models were solved using SAS 9.4 mixed procedure. Specifically, DAT1 genotype, OPRM1 genotype, their interaction terms and covariates were all included in the models; we tested whether DAT1 and OPRM1 genotype independently or in epistatic interaction predicted subjective response scores on the SHAS, BAES stimulation and sedation subscales, DEQ VAS positive and negative effects subscales, and TLFB drinking measures (number of drinking days, number heavy drinking days and average drinks per drinking day). We ran full diagnosis on all the statistical models included in this manuscript to make sure all mathematical assumptions for the models were met and the models were valid. Significance level was set at P= 0.05, uncorrected. We were testing multiple correlated hypotheses, under carefully controlled testing conditions and number of subjects tested; given our sample size, we felt a Bonferroni correction would be too conservative for this exploratory analysis.
