Sum scores were tested for association with DSM-IV AD in the case-control COGA and SAGE samples using logistic regression with sex as a covariate in COGA and sex, age quartiles, and study site as covariates in SAGE. The models were selected to follow the methods used in the previously reported primary COGA and SAGE GWAS analyses (Bierut et al., 2010; Edenberg et al., 2010). In addition to testing the sum scores, the individual SNPs contributing to the scores were also tested for association with AD in the sample used for prediction. All association analyses were completed in the case-control samples using logistic regression using an additive model in PLINK v1.07 for both the EA subset of the sample and the entire sample, including individuals of non-EA ancestry. Association analyses in the entire GWAS samples that included individuals of non-EA ancestry included molecularly derived principal components factor covariates, PC1 and PC2, distinguishing primarily between European and African ancestry.
