In this study, we used data from GWAS repositories to estimate the magnitude of imputation-induced bias in a common set of SNPs among European Americans and African Americans genotyped on different Illumina and Affymetrix arrays. We hypothesize that using differing sets of genotyped SNPs from the different arrays as inputs creates differential imputation accuracy across samples resulting in the bias and spurious associations observed by others (Sinnott and Kraft 2012; Uh et al. 2012). However, in contrast to these studies which imputed each sample separately based on their differing sets of genotyped SNPs and then combined the imputed data for analysis (Sinnott and Kraft 2012; Uh et al. 2012) or used imputed data for some samples and genotyped data for other samples (Uh et al. 2012), we tested an intersection strategy in which we selected only the SNPs genotyped on all arrays for the samples to be combined and then imputed up to a common set of HapMap SNPs for analyses from a common set of genotyped SNPs. To test this hypothesis and correction strategy, we examined the degree to
