Figure 3

Power ( y -axis) as a function of the total number, M , of SNPs. Figure 3 shows that power is strongly reduced by the presence of random (unassociated) markers. For 100 cases and controls each, under a model with main and interaction effects (a₁ = 1, a₂ = 0, a₃ = 6), pattern approach (solid red line) and single-locus approach (broken blue line) suffer approximately the same power losses.