To obtain a more meaningful measure of the significance of our smallest P-values, we used the method of (Li and Ji, 2005) to determine that there were 75 independent tests of association after factoring in the correlations among SNPs. In addition, there was a 50% correlation between the phenotypes tested. Thus, we adjust our significance threshold based on 113 independent tests (75 SNPs times 1.5 phenotypes), yielding an experiment-wide P-value cutoff of 0.0004. Only two P-values, for the associations between rs1229984 and DSM-IV symptom count (P = 0.0003), and the same SNP with max drinks (P = 0.0004), exceed this threshold. The smallest P-values for SNP x SNP interactions are on the order of 2.0 × 10−5, which are unimpressive P-values given the number of pairwise interaction tests performed.
