1# Verification checks (P4)2| # | check | outcome |3|---|---|---|4| 1 | Delete results.json, re-run analysis.py; SHA-256 identical (e4686ab7…955c1b both runs) | PASS |5| 2 | unverifiable_references == refs failing >=1 of the three checks | PASS |6| 3 | n_verified (4) + len(unverifiable) (1) == 5 == n_analysed; exclusions: 5 − 0 = 5 | PASS |7| 4 | n_input_rows equals `wc -l` counts (README 17, sources 134/58/107/66/91) and JSON item count 5 | PASS |8| 5 | Source SHA-256 equal README values; data/ unmodified after runs | PASS |9| 6 | Every locator names source id and table/section | PASS |10| 7 | Year rule applied: R4 entry 2025 = v1 year (v3 2026) accepted; R3 2024 ∉ {2026, 2026} | PASS |11| 8 | Arithmetic: exp(−0.431)−1 = −35.0% (R1); −0.0962/0.261 = −0.3686 vs printed −0.3685 (printed-value rounding, consistent with 36.85%) (R5); −0.268/0.366 = −0.732 (R4) | PASS |12| 9 | Baseline (title+year only) flags {R3}; full check also flags only {R3} — consistent, finding check adds no extra flag | PASS (noted) |13| 10 | Source-internal inconsistencies (2609 SE 0.210 vs 0.218, 608 vs 640; 2505 SE 0.091 plausibility) | Reported in data_issues, not silently ignored |14No randomness; random_seed null.15