Fig 1: Violin plot of FDP for varying numbers of multiple tests, correlation 0, from simulated data. Test stat dist's under H0 and HA are unit normal with means 0 and $0.8 \sqrt{n/2}$, respectively, corresponding to a paired 2 sample test, effect size $0.8$. Prior prob that HA is true is 0.05, with sample size, 45, giving 85\% average power. FDP under BH-FDR procedure, FDR=15\%.

Fig 1: Violin plot of FDP for varying numbers of multiple tests, correlation 0, from simulated data. Test stat dist’s under H0 and HA are unit normal with means 0 and \(0.8 \sqrt{n/2}\), respectively, corresponding to a paired 2 sample test, effect size \(0.8\). Prior prob that HA is true is 0.05, with sample size, 45, giving 85% average power. FDP under BH-FDR procedure, FDR=15%.

\[\\[1in]\]

Fig 2: Violin plot of FDP for varying numbers of multiple tests, correlation 0.50, from simulated data. Test stat dist's under H0 and HA are unit normal with means 0 and $0.8 \sqrt{n/2}$, respectively, corresponding to a paired 2 sample test, effect size $0.8$. Prior prob that HA is true is 0.05, with sample size, 45, giving 85\% average power. FDP under BH-FDR procedure, FDR=15\%.

Fig 2: Violin plot of FDP for varying numbers of multiple tests, correlation 0.50, from simulated data. Test stat dist’s under H0 and HA are unit normal with means 0 and \(0.8 \sqrt{n/2}\), respectively, corresponding to a paired 2 sample test, effect size \(0.8\). Prior prob that HA is true is 0.05, with sample size, 45, giving 85% average power. FDP under BH-FDR procedure, FDR=15%.

\[\\[1in]\]

Fig 3: Violin plot of TPP for varying numbers of multiple tests, correlation 0, from simulated data. Test stat dist's under H0 and HA are unit normal with means 0 and $0.8 \sqrt{n/2}$, respectively, corresponding to a paired 2 sample test, effect size $0.8$. Prior prob that HA is true is 0.05, with sample size, 45, giving 85\% average power. TPP under BH-FDR procedure, FDR=15\%.

Fig 3: Violin plot of TPP for varying numbers of multiple tests, correlation 0, from simulated data. Test stat dist’s under H0 and HA are unit normal with means 0 and \(0.8 \sqrt{n/2}\), respectively, corresponding to a paired 2 sample test, effect size \(0.8\). Prior prob that HA is true is 0.05, with sample size, 45, giving 85% average power. TPP under BH-FDR procedure, FDR=15%.

\[\\[1in]\]

Fig 4: Violin plot of TPP for varying numbers of multiple tests, correlation 0.50, from simulated data. Test stat dist's under H0 and HA are unit normal with means 0 and $0.8 \sqrt{n/2}$, respectively, corresponding to a paired 2 sample test, effect size $0.8$. Prior prob that HA is true is 0.05, with sample size, 45, giving 85\% average power. TPP under BH-FDR procedure, FDR=15\%.

Fig 4: Violin plot of TPP for varying numbers of multiple tests, correlation 0.50, from simulated data. Test stat dist’s under H0 and HA are unit normal with means 0 and \(0.8 \sqrt{n/2}\), respectively, corresponding to a paired 2 sample test, effect size \(0.8\). Prior prob that HA is true is 0.05, with sample size, 45, giving 85% average power. TPP under BH-FDR procedure, FDR=15%.