A second method for using IQS without directly genotyping would be to develop a database of common imputations in common populations that records IQS scores for each SNP. To test the practicality of this approach, we randomly divided the data into two groups and tested the robustness of the IQS score for the SNPs imputed from the Illumina 550 K array to the Illumina 1 M array in both EA and AA. Because small changes in the denominator of IQS (1-Pc) will dramatically affect the value of the statistic when MAF is small, we included only SNPs with MAF>0.01. Figure 3 plots the IQS scores in both populations. The correlation in EA is 0.99519 and the correlation in AA is 0.99020, indicating that IQS is robust for the same imputation in a relatively homogeneous population.
