Skip to content

fix(calibrate): disclose Kendall-tau reuses kappa's Landis-Koch cutoffs#1

Merged
Mike-E-Log merged 1 commit into
mainfrom
fix/tau-band-honesty
May 20, 2026
Merged

fix(calibrate): disclose Kendall-tau reuses kappa's Landis-Koch cutoffs#1
Mike-E-Log merged 1 commit into
mainfrom
fix/tau-band-honesty

Conversation

@Mike-E-Log
Copy link
Copy Markdown
Owner

Relabels the tau band_basis to match the code: tau has no established band standard, so it reuses kappa's Landis-Koch cutoffs as a
disclosed heuristic. Tests: 43 passed, 1 skipped.

The τ band_basis claimed "heuristic strength bands (NOT Landis-Koch)",
but the band was computed by the same landis_koch_band() that κ uses — the
exact Landis-Koch cutoffs. The label denied what the code did, contradicting
the repo's own no-over-claiming thesis.

Fix the disclosure to match the code rather than invent a second arbitrary
τ scale (which would be less honest, not more): τ has no established band
standard, so the tool reuses κ's cutoffs as a DISCLOSED heuristic and says so.

- rename landis_koch_band() -> strength_band() (metric-agnostic; docstring
  explains κ=standard vs τ=disclosed-heuristic)
- τ band_basis now: "tool heuristic: reuses the Landis-Koch kappa cutoffs
  (tau has no established band standard, so treat the label as rough)"
- flip test_parity's τ test from pinning the false "NOT Landis-Koch" claim
  to pinning the honest reuse disclosure
- sync README + CHANGELOG wording

Tests: 43 passed, 1 skipped.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@Mike-E-Log Mike-E-Log merged commit 66384d6 into main May 20, 2026
2 checks passed
@Mike-E-Log Mike-E-Log deleted the fix/tau-band-honesty branch May 20, 2026 23:58
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant