Figure 2
From: A novel approach to minimize false discovery rate in genome-wide data analysis
Simulation result I. Compare the average performance of BH, Storey, SAM and miFDR on 1000 simulation runs. In each plot, a blow-out of the curve segment in the dashed rectangle is shown at the bottom-right corner for clearer illustration. (a) Compare the estimated FDRs among four approaches. The performance of miFDR is the best. (b) Compare the true FDRs. The performance of miFDR is the best. (c) Compare the estimated FDRs and the true FDRs of miFDR. The true FDR curve of miFDR is well bounded by its estimated FDR curve, which indicates that miFDR does not under-estimate the number of falsely called significant features.