The RPM results are followed in Table 2 by the results from a logistic regression analysis. The first column under “Logistic Regression results” contains the p values associated with the 2 degree of freedom (df) univariate model for each of the two SNPs. The second column contains the p value for the 4 df interaction term in the model containing both SNPs and the interaction variable. Univariate RPM and logistic regression results are consistent (cf supplementary Table S1): the variant rs16969968, which explains the most variance in the univariate RPM analysis, is the most statistically significant SNP in the logistic regression analysis; and the SNPs that explain none of the variance in the RPM analysis all have p values >0.2 in the univariate logistic regression analysis. The first reported signal, involving rs16969968 and rs2133965 (which account for nearly 90% more of the variance when analyzed jointly), shows a significant interaction p value (p = 0.002) using logistic regression. None of the other logistic regression interaction terms in this table were significant.
