Data audit (P1)
| file | lines (wc -l) | SHA-256 | matches README/bundle |
|---|---|---|---|
| data/README.md | 10 | 16ebdff389eeaee9508fe81c4edc43690f31f3463c469680c814134086ef9929 | matches bundle.json |
| data/sources/arxiv-2609.27732v1.txt | 134 | ee2d444d2f78f2af2ca2cf69b59a61fcd7ac5e10ea29e8ddcd67917bf07d6a8b | yes |
| data/sources/arxiv-2606.14621v1.txt | 58 | 27c463c0a144d9d5a3916dc641f15b6029f6c68b5e574ad9074e7df1aeb0a238 | yes |
Schema: plain text; tables one row per line with " | " separators. DCR tables: coefficient row followed (after a blank line) by an SE row in parentheses; columns 1 km / 3 km / 5 km. Units: log points (ATT), counts (regions, region-years). RESEA Table 2 (re-set by extract editor): 5 columns, cells hold "estimate; p=..; 95% CI: [lo, hi]"; USD/yr, pp, weeks. No dates beyond submission/retrieval (2026). No missing cells in the requested rows; no duplicates.
Anomalies / internal inconsistencies (handled in results.json data_issues)
- DCR 5 km CS-DR ATT -0.431 has SE (0.210) in Table 6 but (0.218) in Table A19. Keys name Table 6 -> use 0.210.
- Table A19 note says 640 never-treated comparison regions; Tables 6/7 and Figure 6 note say 608. Not used for any key; reported.
- Abstract says fatal hospitalisations "fall by 35%"; Section 6.1 says "approximately 30% at 5 km"; Table 7 5 km fatal ATT -0.354 => exp(-0.354)-1 ~ -29.8%. Not used; reported.
- Text says "HIV estimates are insignificant and close to zero" though Table 7 5 km HIV ATT is -0.340 (SE 0.294); insignificance is consistent with no star; "close to zero" is the author's wording. Reported only.
- Figure 7 note / Table 7 note list four sub-categories while Table 7 has five outcome rows. Reported only.
- RESEA text reports the weeks-on-UI CI as positive magnitudes [1.75, 2.24] (a reduction); Table 2 prints [-2.24, -1.75]. Keys use Table 2.
- RESEA text contains a rendering defect "p ¡ 0.001" (should read p < 0.001). No key affected.
- RESEA wage ATE CI [166, 2,141] is near-symmetric around 1,153 (987 vs 988), consistent with rounding.
- The exact two-sided p-value of the HIV 5 km ATT is not printed anywhere (only star convention) -> null.
Availability
Both extracts are admissible evidence; nothing violates the rules. README.md read for provenance only. RESEA Table 3 and DCR Table A19 duplicate some values; keys name Tables 2 and 6, which are used.
Exclusions
No rows excluded; the analysis units are the 25 requested reported items (see out/exclusions.csv).