The difference between the pre- and postanalysis information measures, MACH r2 and RT2, is shown in Figure 2. Using Sib 1 and controls data, we randomly selected 1000 SNPs each from three classes of SNPs: P-values >greater than 0.05, P-values smaller than 0.001, and intermediate ones. Although for unassociated SNPs (P-value >0.05) the two measures show good agreement, they are quite different for strongly associated SNPs (P-value <0.001). The post-analysis measure, therefore, can be a useful tool for selecting SNPs for meta-analysis.
