# Verification checks (P4)

• Clean re-run of out/analysis.py: results.json byte-identical (cmp) -> PASS
• all SEs positive -> PASS
• DCR CI contains estimate -> PASS
• DCR CI symmetric half-width = 1.96*SE -> PASS
• DCR CI excludes 0 consistent with ** star -> PASS
• RI wage CI contains estimate -> PASS
• RI weeks CI contains estimate and lo<hi -> PASS
• RI wage CI ~symmetric (midpoint within 1 of estimate) -> PASS
• normal p from derived SE rounds to printed 0.022 -> PASS
• no star mismatches across 22 DCR cells -> PASS
• HIV 5km not significant (no star) and normal p>0.10 -> PASS
• p_value null with 'not reported' locator -> PASS
• dcr_obs_5km integer -> PASS
• locators: 23 keys (all but 2 booleans) -> PASS
• n_input_rows matches wc -l (10/134/58) -> PASS
• exclusion ledger: 25-0=25=n_analysed -> PASS
• ATE magnitude vs control mean 3.5% (text) -> PASS