Figure 2.

FDR control and power comparison of different methods in simulation studies. $m=10\hspace{0.17em}000,\hspace{0.17em}{\xi }_{11}=0.05$ and ${\xi }_{01}={\xi }_{10}$. Each row corresponds to different ${\xi }_{00}$. Each column corresponds to different standard deviations. In each panel, the empirical FDR for different methods was calculated over 100 replications at a target FDR level of 0.05 (horizontal dashed line in the plots) for different non-null settings (left: ${\mu }_1=2.5,{\mu }_2=3$; middle: ${\mu }_1={\mu }_2=2.5$; right: ${\mu }_1={\mu }_2=3$).