Compatibility / confidence curve = the two-sided Wald p-value function across candidate effect sizes. Relative likelihood = a monotone transform of the same Wald distance, normalized to 1 at the CI-implied estimate. S−2 interval = candidate effects with support no less than −2 versus the MLE, equivalent here to relative likelihood at least exp(−2) or the MLE being no more than 7.4x as supported.

Both views come from the same CI-based Wald reconstruction. The likelihood view is often easier to read as relative evidentiary support across effect sizes under that Wald approximation. This app does not reconstruct the exact fitted-model profile likelihood; likelihood-principle-style interpretation applies within the Wald reconstruction shown here, not as a claim that the original model likelihood has been recovered.

The paired 80% power markers are a legacy closed-form benchmark at alpha 0.05 and target probability 0.80. They are not the exact critical-effect solution and are not an MCID.