Logistic regressions were performed on all CNV regions. After identifying potential regions, individual dummy variables for duplications and deletions were created to dissect the association signal with DSM-IV alcohol dependence. Several covariates were included in the model based on the previous GWAS of these data (Bierut et al., 2010), including sex, age, and two principal components indexing continuous ancestral genetic variation. We also included a dummy variable to indicate the source of DNA (cell line versus whole blood). In addition, we ensured that these potentially confounding variables were not directly associated with the identified CNVs.
