The procedure for set-based testing in PLINK was modified and adapted for these analyses3 (Perlis et al., 2008; Purcell et al., 2007). In the modified set-based approach used here, a null distribution was formed via 100,000 label-swapping permutations, in which the outcome phenotype (dependence symptoms), CSA score, and covariates of no interest were permuted together. A participant's full genome was left intact to preserve the linkage structure across individual SNPs during permutations.
