Table 2. Comparing results from the three approaches to estimate and detect a trend across the different scenarios. α is the threshold for statistical significance.
| Unregularized | Informative | Weakly-informative | |
|---|---|---|---|
| Unconditional on statistical significance | |||
| Bias | over- or under-estimation | underestimation | |
| Coverage | includes the true value less than 1 − α times | ||
| Conditional on statistical significance | |||
| Power | <80% except for large decline | ||
| Coverage | includes the true value less than 1 − α times | ||
| Type-I error | inflated (>α) | under control (≤α) | |
| Type-S error | large for small declines | ||
| Type-M error | underestimation | over- or under-estimation | underestimation |