False discovery rate procedures for high-dimensional data
KI Kyung In Kim · TU/e Research Portal · 2008
Tables 1.1 Illustration of possible configuration for m hypothesis testing.U , V , T , S, W and R are all random variables representing the number of hypotheses falling on each category.m 0 and m 1 are the number of true and false null hypotheses respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.1 Comparison of FDR for six different error distributions.Average FDR, standard deviation of FDR, average threshold, average number of bins and average computing time (minutes:seconds) are shown, based on ten simulations. . . .26 2.2 Comparison of robustness removing one slide in turn. . . . . . . . . . . . . . .29 3.1 Simulation results for cases with correlation parameter ρ equal to 0, 0.2, 0.4, 0.6 and 0.8 when m = 3000, m 0 = 2000 and m 1 = 1000.Data are generated from the procedures in Section 4 of Qiu and Yakovlev (2007). . . . . . . . . .38 4.1 Median number of detected genes under increasing edge densities and the corresponding correlation matrices . . . . . . . . . . . . . . . . . . . . . . . . . .50 4.2 16 genes showing different significance feature under nested 10 correlation matrices . . . . . . . . . . . . .