The genotype distributions for some SNPs may differ markedly across platforms due to genotyping artifact or differences in imputation quality. These differences may be identified even in relatively small samples. We explored the possibility of genotyping a small number of additional controls along with the cases, which could be used to identify and eliminate the problematic SNPs. Researchers would perform a preliminary analysis comparing the additional controls to the original controls, and any SNP significant in this preliminary analysis would be discarded. Researchers could then perform standard association tests between cases and controls using the remaining SNPs.
