Figure 1
From: A comparison of internal validation techniques for multifactor dimensionality reduction
Overview of the MDR method for k = 2 loci. First all possible combinations of k = 2 loci are enumerated. For a given combination of loci, the number of cases and controls are tabulated for each genotype combination, and then the ratio of cases to controls is calculated within each cell. If the ratio exceeds a threshold (1.0 here) then the combination is labeled as high risk, otherwise it is labeled as low risk. This high-risk/low-risk characterization is the MDR model for that combination of loci, and the accuracy of the model is determined. The model that maximizes accuracy is chosen as the best model of size k. Repeat the process for a new k.