Table 6 Well-approximated type I error rates and power with weak informative prior for variance of BHM
From: Practical basket design for binary outcomes with control of family-wise error rate
| Â | Â | Â | Subpopulation j | |||
|---|---|---|---|---|---|---|
Scenario | Method | \(\tilde{\sigma }^2\) | 1 | 2 | 3 | 4 |
1 | BHM-weak | \(\tilde{\sigma }^2 = 25\) | 1% | 1% | 1% | 1% |
\(\tilde{\sigma }^2 = 50\) | 1% | 1% | 1% | 1% | ||
BHM-strong | \(\tilde{\sigma }^2 = 25\) | 1% | 1% | 2% | 1% | |
\(\tilde{\sigma }^2 = 50\) | 1% | 1% | 2% | 1% | ||
2 | BHM-weak | \(\tilde{\sigma }^2 = 25\) | 1% | 1% | 1% | 66% |
\(\tilde{\sigma }^2 = 50\) | 0% | 1% | 1% | 66% | ||
BHM-strong | \(\tilde{\sigma }^2 = 25\) | 2% | 4% | 5% | 70% | |
\(\tilde{\sigma }^2 = 50\) | 2% | 3% | 4% | 71% | ||
3 | BHM-weak | \(\tilde{\sigma }^2 = 25\) | 1% | 1% | 73% | 67% |
\(\tilde{\sigma }^2 = 50\) | 1% | 1% | 72% | 67% | ||
BHM-strong | \(\tilde{\sigma }^2 = 25\) | 6% | 8% | 85% | 78% | |
\(\tilde{\sigma }^2 = 50\) | 6% | 8% | 85% | 78% | ||
4 | BHM-weak | \(\tilde{\sigma }^2 = 25\) | 1% | 71% | 74% | 68% |
\(\tilde{\sigma }^2 = 50\) | 1% | 71% | 74% | 67% | ||
BHM-strong | \(\tilde{\sigma }^2 = 25\) | 15% | 87% | 93% | 87% | |
\(\tilde{\sigma }^2 = 50\) | 15% | 87% | 92% | 87% | ||
5 | BHM-weak | \(\tilde{\sigma }^2 = 25\) | 72% | 71% | 75% | 69% |
\(\tilde{\sigma }^2 = 50\) | 72% | 71% | 74% | 68% | ||
BHM-strong | \(\tilde{\sigma }^2 = 25\) | 93% | 93% | 96% | 92% | |
\(\tilde{\sigma }^2 = 50\) | 93% | 92% | 96% | 92% | ||