Principal components analyses (PCA) were conducted with data from cases and ATR controls using the smartpca program in the Eigensoft 3.0 package (Patterson, Price, & Reich 2006). Due to the dense coverage of high priority candidate genes, the kill r2 option was set at 0.8 for these analyses. Our inclusion of data from AIMs in these analyses prevented using Tracy-Widom statistics to determine the number of principal components (PCs) to be used as covariates. As such, we opted to include all PCs for which case-control differences reached at least a trend level of significance. Post-hoc analyses which included a larger number of PCs (ten) as covariates were performed to demonstrate that our primary findings remained significant despite this additional control for admixture (see limitations).
