# Checks (P4) — 2026-10-30

• Clean re-run of out/analysis.py (outputs deleted first): results.json and analytic.csv byte-identical: PASS
• n_valid+n_unreadable==control: PASS
• n_analysed==n_valid: PASS
• analytic rows==n_analysed: PASS
• analytic price sum==sum_price_valid: PASS
• implied sum identity: PASS
• mean==sum/n: PASS
• median ppm2 recomputed: PASS
• exclusions chain: PASS
• ids unique: PASS
• muni shortfall sums to n_unreadable: PASS
• implied mean of missing plausible (between min,max of valid): PASS
• feasibility partial since n_unreadable>0: PASS
• Magnitudes: baseline control sum 471,287,000 vs valid 456,370,000 (96.8%); valid count 1164/1200 (97.0%); consistent.