Our investigation into the nature of the bias observed under the union-of-SNPs imputation strategy showed minimal overlap in the SNPs with spurious association between the two ancestry groups. Additionally, making post-imputation SNP exclusions for R2 < 0.3 removed some, but not all of the spurious SNP associations (Figure S5). After the R2 exclusion, the remaining SNPs with spurious association tended to have MAF <10 % (especially in European Americans) and/or large discrepancies in R2 between the two studies (especially in African Americans).
