Figure 1. .

Summary of the steps involved in implementing the MDR and GMDR methods (adapted from the work of Ritchie et al.¹⁴ and Hahn et al.¹⁵). The two methods share the same reduction strategy. The difference is that, in the GMDR method, we substitute a score statistic or some other quantitative measure, instead of affection status, to define the two different groups. In balanced case-control studies with no covariate, the two methods are exactly equivalent—that is, given an equivalent threshold, the two methods yield the same results, including the best model and classification and prediction accuracies. For a detailed description of the steps, please see the “Score-Based MDR Method” subsection. In step 3, bars represent hypothetical distributions of cases (left, dark shading) and controls (right, light shading); numbers not in parentheses above bars are the numbers of cases and controls, and those in parentheses are the sums of the scores. In steps 4 and 6, numbers not in parentheses are the ratios of the number of cases to the number of controls, and those in parentheses are the average scores. “High-risk” cells are indicated by dark shading, “low-risk” cells by light shading, and “empty” cells by no shading.