As would be expected a priori, there were some inconsistencies in the genotypes of the 10,760 SNPs that were successfully imputed using the separate ancestral reference panels (CEU and YRI). Obviously, this issue needs resolution before generated datasets can be merged. The simplest protocol to address this situation is to preferentially accept the imputed genotypes resulting from the reference panel giving the highest concordance, which for our data would lead to a preference of calls based on the phase II YRI reference panel over calls based on the phase II CEU reference panel.
