A second notable result is that the quality of imputation in AA is markedly lower than in EA. This is seen in the decreased efficiency by nearly ten percentage points, and decrease in mean IQS by nearly twelve percentage points. This is likely due to two factors. First, African Americans have more diverse haplotypes and more uncommon alleles. Second, there is non-negligible difference between African Americans and the YRI reference panel, which was clearly reflected by Eigenstrat population structure analysis [17]. Interestingly, imputation accuracy is nearly the same for EA as for AA, again highlighting how imputation accuracy can overestimate the quality of imputation.
