Figure 3.—

Power of RWAS with different prior information. For each group PAR, 10,000 data sets were generated, and each data set contained 1000 case and 1000 control individuals having 100 variants with predefined true c_i values. c_i of 50 variants was 0.8, and c_i of the other 50 variants was 0.2. Five different types of prior information were given to RWAS: “correct c_i” (same c_i as true c_i of data sets), “uniform incorrect c_i” (c_i = 1 for all variants), “three-fourths correct c_i” (three-fourths of true c_i), “one-half correct c_i” (one-half of true c_i), and “very incorrect c_i” (opposite c_i to true c_i of data sets). The single-marker test, MB, and RWAS with the five different types of prior information were tested.