| 1 |
Clean re-run: deleted out/results.json, re-ran analysis.py; SHA-256 identical across runs (edb19ad1…cd8c) |
pass |
| 2 |
Counts identity: n_supported + n_not_supported + n_not_assessable = 2+3+1 = 6 = n_analysed = number of assertions |
pass |
| 3 |
Exclusion ledger: n_raw 6 − removed 0 = n_analysed 6 |
pass |
| 4 |
Every verdict in allowed set; one locator and one reason per id A1–A6 |
pass (asserted in script) |
| 5 |
Locators point to the source each assertion cites (A1–A5 → arxiv-2609.27732v1, A6 → arxiv-2606.14621v1) |
pass |
| 6 |
Log-to-% conversions reproduce the source's printed % rows: −0.431→−35.0%, −0.283→−24.7% |
pass |
| 7 |
A1 significance: |
−0.431 |
| 8 |
A4 CI with 1.96 [−0.896, 0.084] and 2.0 [−0.906, 0.094] both contain 0; CI contains estimate |
pass |
| 9 |
A6: 1.99/17.3 = 11.50%; CI [−2.24, −1.75] contains −1.99 |
pass |
| 10 |
A5: regex search for men/women/male/female/sex/gender in DCR extract → 0 hits |
pass |
| 11 |
Input hashes match README/bundle |
pass |
| 12 |
Baseline (number-match) vs final: baseline flags A3, A4 as having unmatched numbers (agrees with not_supported); baseline misses A2 (sign error with matching digits) and cannot detect A5 — expected limitation of naive baseline |
informative |