Table 3.
Summary of Type I error simulation with N = 216 and balanced allocation of subjects to B groups. Results are displayed as the number (%) of cells in simulation experiment where type I error was significantly below or above the nominal 5% based on simulation error bounds, and the range of observed type I error rates.
| Model | Below | Above | Type I Error Range | |
|---|---|---|---|---|
| nB = 6 | AB (KR nobound) | 13 (41%) | 16 (50%) | 1.8–13.1 |
| AB (KR bound) | 25 (78%) | 0 (0%) | 1.4–4.5 | |
| A (KR nobound) | 1 (3%) | 13 (41%) | 3.5–27.0 | |
| A (emp) | 0 (0%) | 31 (97%) | 6.3–36.5 | |
| B (KR nobound) | 1 (3%) | 14 (44%) | 3.6–27.5 | |
| B (emp) | 0 (0%) | 32 (100%) | 6.9–36.3 | |
| IND | 0 (0%) | 24 (75%) | 4.6–53.3 | |
| nB = 18 | AB (KR nobound) | 17 (35%) | 0 (0%) | 2.7–6.1 |
| AB (KR bound) | 30 (63%) | 1 (2%) | 1.4–6.4 | |
| A (KR nobound) | 0 (0%) | 9 (19%) | 4.0–15.2 | |
| A (emp) | 0 (0%) | 48 (100%) | 6.5–22.0 | |
| B (KR nobound) | 1 (2%) | 24 (50%) | 3.5–35.5 | |
| B (emp) | 0 (0%) | 32 (67%) | 5.2–38.5 | |
| IND | 0 (0%) | 33 (69%) | 4.6–42.7 | |
| nB = 36 | AB (KR nobound) | 17 (35%) | 2 (4%) | 2.3–6.5 |
| AB (KR bound) | 27 (56%) | 1 (2%) | 1.5–6.5 | |
| A (KR nobound) | 1 (2%) | 10 (21%) | 3.3–12.3 | |
| A (emp) | 0 (0%) | 48 (100%) | 7.0–18.1 | |
| B (KR nobound) | 1 (2%) | 25 (52%) | 3.5–38.1 | |
| B (emp) | 0 (0%) | 28 (58%) | 4.3–39.4 | |
| IND | 0 (0%) | 33 (69%) | 4.5–39.9 | |
| nB = 72 | AB (KR nobound) | 18 (38%) | 1 (2%) | 2.2–6.7 |
| AB (KR bound) | 23 (48%) | 1 (2%) | 1.9–7.0 | |
| A (KR nobound) | 1 (2%) | 6 (13%) | 3.6–7.6 | |
| A (emp) | 0 (0%) | 48 (100%) | 6.5–12.2 | |
| B (KR nobound) | 0 (0%) | 25 (52%) | 3.7–39.7 | |
| B (emp) | 0 (0%) | 25 (52%) | 4.0–40.1 | |
| IND | 0 (0%) | 31 (65%) | 4.1–38.7 | |