Skip to content

FIX-7-1: pre-registration for the threshold calibration study #416

Description

@mateuszwalo

Track: research · Level: core · Effort: ~6h · Phase: 7 · Depends on: FIX-3-2
Source: this plan, section 9

Why this matters

T20's credibility comes from having frozen the statistical contract before looking at the
numbers. The calibration study has more degrees of freedom than T20 did, so it needs the same
discipline or it will find a threshold that fits and prove nothing.

Steps

  1. Write down, before any calibration run: the decision rule, the threshold grid, the loss
    quantile grid, the train/validation seed split, the primary endpoint, and the statistical
    contract.
  2. State the falsifiable prediction, in the T20 style.
  3. State the kill criterion explicitly: if the calibrated rule does not beat metric_argmax
    on held-out seeds, diagnosis_single does not become the default and the result is published
    as a negative result.
  4. Get maintainer sign-off before the first run.

Done when

The pre-registration document is merged and dated before any calibration output exists.


How to take this: comment "taking this" and wait to be assigned. Branch fix-7-1-preregistration from upstream/main, and put Closes #<this issue number> in your PR. Full workflow: the Cohort Handbook (pinned in Discord).

Plan: plans/FIX-PLAN-T20-2026-08-23.md, section 9.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    P1Audit backlog: before next releasecohortBNNR 2026 cohort tasklevel:coreCore tasktrack:researchCohort research task

    Type

    No type

    Projects

    No projects

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions