Timing methods and evidence
Compact methods for the plan-frozen TIMING-ECO-3 retrospective. Every public number on this page is taken from the append-only adjudication ledger.
Protocol
- Completed post-2002 seasons only; the pre-2002 net regime does not enter the panel.
- Target: observed-index P50 as day of season, with 10 June equal to zero.
- Non-fishing days stay operationally missing. They are never biological zeroes.
- Rolling-origin evaluation with training years before each test year.
- Seven prespecified Holm-family predictors. Family size 7.
- A candidate becomes a retrospective champion only if it passes frozen predictive-usefulness gates. None did.
Reviewed result
- Holm-adjusted p-values below 0.05: 0.
- Smallest Holm-adjusted p-value: 0.06887862242755145 (air_canhomt_mean_c).
- Predictive usefulness passed: false.
- Retrospective champion selected: false.
- Public prediction ready: false.
- Adjudication record: tec3-adj-177d97a2f3f43938490885e35c40fdcf.
observed_index_association_ready is true only as readiness to report the reviewed descriptive observed-index association results. None of the seven Holm-adjusted p-values is below 0.05; the smallest is the CanHomT air result at 0.06887862242755145. Readiness is not a claim that a statistically adjusted association exists. observed_index_prediction_ready remains false because predictive_usefulness_passed is false and no champion was selected. The modeled target remains observed-index P50.
The generated TIMING-ECO-3 readiness packet remains the immutable pre-review record and stays locked. Public reporting uses the append-only adjudication ledger. Association reporting may be described as ready only because that ledger says so. That flag is report-readiness, not a statistically adjusted association claim, not prediction readiness, and not a selected champion.
The completion tracker on the home page is a different target and a different model. This retrospective does not change the issued champion, baseline, or descriptive completion range.