Each of the joint effect signals in Table 2 involves the SNP rs16969968. This SNP provides the strongest signal in the univariate RPM analysis of the COGEND data, accounting for 1.22% of the trait variance. Because selection focused on maximizing explained variance, it is not surprising to find rs16969968 in each of the signals. We note that RPM models for two of the secondary SNPs accounted for none of the trait variance in the single SNP analyses; however, when jointly analyzed with rs16969968, additional trait variance is explained.
