Data audit
| file | SHA-256 | lines (wc -l) / items | matches README/bundle |
|---|---|---|---|
| data/assertions.json | 2965ab3acc6f326653169b83a5345698b51fac88a556de65faf89210aea0cad4 | 6 items (JSON list) | yes (bundle) |
| data/README.md | 5348bd49912099735108dbbe19ae09fa72bd600cb3d12b878658e6b9806198a7 | 14 lines | yes (bundle) |
| data/sources/arxiv-2609.27732v1.txt | ee2d444d2f78f2af2ca2cf69b59a61fcd7ac5e10ea29e8ddcd67917bf07d6a8b | 134 lines | yes |
| data/sources/arxiv-2606.14621v1.txt | 27c463c0a144d9d5a3916dc641f15b6029f6c68b5e574ad9074e7df1aeb0a238 | 58 lines | yes |
Schema: assertions = list of {id, cites, text}; 6 unique ids A1..A6, no duplicates, no missing fields; A1-A5 cite arxiv-2609.27732v1, A6 cites arxiv-2606.14621v1. Units: DCR = log points (ATT) / % change; RESEA = dollars, pp, weeks.
Anomalies (internal inconsistencies of sources):
- DCR Table A19 CS-DR SE (0.218) differs from Table 6 5 km SE (0.210) for the same ATT -0.431.
- DCR Table A19 note says 640 never-treated regions; table body and Tables 6/7 say 608.
- DCR abstract: fatal outcomes "fall by 35%"; Table 7 5 km fatal ATT -0.354 ⇒ -29.8%, and Sec 6.1 says "approximately 30%".
- DCR text says HIV estimates "close to zero"; Table 7 HIV 5 km is -0.340 (SE 0.294) — insignificant but not small.
- RESEA results text renders "p ¡ 0.001" (encoding defect for "p < 0.001").
- RESEA text gives the weeks-on-UI CI in magnitude form [1.75, 2.24] while Table 2 gives [-2.24, -1.75]; consistent, sign convention only. No input violates the availability rules (all are the admissible extracts). No rows excluded.