TABLE 5.
True discovery rate for the envelope test with partial differences (pro = 40% and pro = 10%)
| Simulation scenario | pro | T 1 | T 2 | T 3 | T 4 | T 5 | T 6 |
|---|---|---|---|---|---|---|---|
| Scenario 2 | pro = 40% | 0.180 | 0.441 | 0.514 | 0.576 | 0.600 | 0.151 |
| pro = 10% | 0.088 | 0.343 | 0.440 | 0.526 | 0.565 | 0.152 | |
| Scenario 3 | pro = 40% | 0.554 | 0.959 | 0.991 | 0.994 | 0.999 | 0.042 |
| pro = 10% | 0.336 | 0.896 | 0.975 | 0.992 | 0.994 | 0.053 | |
| Scenario 4 | pro = 40% | 0.761 | 0.998 | 1.000 | 1.00 0 | 1.000 | 0.103 |
| pro = 10% | 0.526 | 0.981 | 0.998 | 1.000 | 1.000 | 0.122 | |
| Scenario 5 | pro = 40% | 0.844 | 0.998 | 1.000 | 1.000 | 1.000 | 0.318 |
| pro = 10% | 0.624 | 0.978 | 1.000 | 1.000 | 1.000 | 0.293 |
Notes: Results for the different scenarios analogous to Table 2 and 3, where the differences between control and test groups are in a random interval of length pro = 40% and pro = 10% of the domain. In all scenarios we are assuming a 5% outlier rate. Numbers represent true discovery rate defined as the proportion of points detected as different with the envelope test among those that are truly different. We excluded Scenario 1 where c = 1 since the two groups are generated from same process and therefore there are no truly different points. Results are based on the envelope test using test statistics T1, T2, T3, T4, T5, T6.