Table 1. .

The Type I error and power for testing the 1,000 random off-diagonal entries in edge correlation at the 0.05 significance level are averaged over 5,000 iterations. The power study is separated into 1,000 random weak signals ∈ (0, 0.2] and 1,000 random moderate signals ∈ (0.2, 0.5], respectively. Simulation results for N = 50 and N = 200 variables over a range of sample size T are reported.

	N = 50			N = 200
	Type I error	Weak signal	Moderate signal	Type I error	Weak signal	Moderate signal
T = 300	0.048	0.237	0.858	0.048	0.163	0.755
T = 500	0.049	0.377	0.971	0.049	0.250	0.922
T = 800	0.049	0.554	0.998	0.049	0.377	0.988
T = 1,000	0.049	0.650	0.999	0.049	0.455	0.997
T = 1,500	0.050	0.813	1	0.049	0.621	1
T = 2,000	0.050	0.904	1	0.050	0.746	1
T = 3,000	0.050	0.972	1	0.050	0.894	1
T = 4,000	0.050	0.989	1	0.050	0.958	1
T = 5,000	0.050	0.994	1	0.050	0.983	1
T = 6,000	0.050	0.997	1	0.050	0.993	1
T = 7,000	0.050	0.998	1	0.050	0.997	1
T = 8,000	0.050	0.999	1	0.050	0.998	1
T = 10,000	0.050	1	1	0.050	0.999	1
T = 15,000	0.050	1	1	0.050	1	1