Table 7: Type I error and power under DGP1.
Method1 |
n0 |
n1 |
ρ1 = −0.4 |
ρ1 = −0.2 |
ρ1 = 0 |
ρ1 = 0.2 |
ρ1 = 0.4 |
PCGC |
500 |
500 |
0.8 |
0.61 |
0.052 |
0.63 |
0.816 |
PCGC |
1000 |
1000 |
0.87 |
0.762 |
0.03 |
0.754 |
0.876 |
PCGC |
1500 |
1500 |
0.892 |
0.804 |
0.044 |
0.79 |
0.89 |
Fork |
500 |
500 |
1 |
0.79 |
0.034 |
0.83 |
1 |
Fork |
1000 |
1000 |
1 |
0.986 |
0.048 |
0.98 |
1 |
Fork |
1500 |
1500 |
1 |
1 |
0.046 |
0.998 |
1 |
TW |
500 |
500 |
0.992 |
0.334 |
0.004 |
0.342 |
0.994 |
TW |
1000 |
1000 |
1 |
0.816 |
0.002 |
0.824 |
1 |
TW |
1500 |
1500 |
1 |
0.98 |
0.004 |
0.974 |
1 |
ρ1 = 0 constitutes the null distribution
1PCGC: Principal Components Group Comparison; Fork: Forkman’s Test; TW: Tracy-Widom