VS activity in each hemisphere were regressed on genotype (coded as 0 vs 1 or more copies of the minor allele) of GWS loci while co-varying for sex, and 3 (AA) or 2 (EA) ancestral principal components using Full Information Maximum Likelihood in MPlus v7.338. Confidence intervals on estimates were derived via bootstrapping (n=10,000). To adjust for multiple comparisons, we used a Bonferroni-corrected p-value threshold (p<0.00625), to account for our hypothesized 8 tests [i.e., rs34066662 and rs75168521 in both brain hemispheres among AAs (4 tests); rs1890881 in both brain hemispheres among AAs and EAs (4 tests)]. As rates of drug dependence, but not alcohol problems (see31), are low in the DNS sample, structural equation models linking genotype to substance dependence with reward-related response as a mediator were not fitted to these data.
