Figure 2: Comparison of performance of T, SAM and the shrinkage statistics that perform best in independent (II) data. Each point represents the areas under the curve for two statistics for a particular simulated data set. In the plots above the diagonal, we report the areas when identifying genes with high signal to noise ratio, while in plots below the diagonal we report areas when identifying genes with large signal. The shrinkage statistics that perform best change across the two goals, and therefore different shrinkage statistics are shown above and below the diagonal. Because over-plotting points may lead to visually misleading results in some cases, we also print the percentages PU and PL of points lying above and below the diagonal, and the averages MX and MY of the horizontal and vertical variables.