# Data audit — F4-conf-01

| file | SHA-256 | lines (wc -l) | matches README/bundle hash |
|---|---|---|---|
| data/README.md | 16ebdff389eeaee9508fe81c4edc43690f31f3463c469680c814134086ef9929 | 10 | yes (bundle.json) |
| data/sources/arxiv-2609.27732v1.txt | ee2d444d2f78f2af2ca2cf69b59a61fcd7ac5e10ea29e8ddcd67917bf07d6a8b | 134 | yes |
| data/sources/arxiv-2606.14621v1.txt | 27c463c0a144d9d5a3916dc641f15b6029f6c68b5e574ad9074e7df1aeb0a238 | 58 | yes |

Schema: plain text; tables rendered one row per line with " | " separators. DCR: SE on the line after each coefficient row, in parentheses; stars ***/**/* = p<0.01/0.05/0.10. RESEA Tables 2–3 re-set to one row per outcome, cells "estimate; p=...; 95% CI: [lo, hi]", thousands separators with commas, "$" prefix.
Units: DCR log points (log hospitalisation rate per 100,000), observations = region-years; RESEA USD/year, pp, weeks.
No duplicates or missing cells in the requested rows. Dates: papers dated 2026-06-12 and 2026-09-23; extracts retrieved 2026-10-30.

## Anomalies / internal inconsistencies
1. DCR: the 5 km CS-DR ATT -0.431 has SE (0.210) in Table 6 but (0.218) in Table A19 (CS-DR column, same 5 km ±5-year spec). Handling: use Table 6 as the keys require; reported in data_issues.
2. DCR: Tables 6/7 report 608 never-treated control regions; Table A19 note says 640 never-treated regions. Not used in any key; reported.
3. DCR: abstract says fatal outcomes "fall by 35%" whereas Section 6.1 says "approximately 30% at 5 km"; exp(-0.354)-1 = -29.8%, so the abstract appears to read log points as percent. Not used; reported.
4. DCR: text says "HIV estimates are insignificant and close to zero" while the 5 km HIV ATT is -0.340 (SE 0.294) — insignificant but not small in magnitude. Not used; noted.
5. DCR: Table 7 prints no exact p-values (only stars) and no Observations row; the exact p-value of HIV 5 km is not reported → null.
6. DCR: Table 7 "Control regions" row rendered as "608 (never-treated) | |" (spanning cell). Rendering artefact, irrelevant to keys.
7. RESEA: text gives the weeks-on-UI CI as positive magnitudes "[1.75 weeks, 2.24 weeks]" whereas Table 2 prints [-2.24, -1.75]. Handling: use Table 2 as printed.
8. RESEA: "p ¡ 0.001" in text (encoding defect for "<"). Irrelevant to keys.
9. RESEA: Table 2 prints no standard errors (only p-values and CIs) → SE derived from CI.
10. Minor % change rounding: Table 6 3 km -24.7% vs exp(-0.283)-1 = -24.65%; A19 -30.3%/-13.9% vs -30.4%/-13.8% — consistent with unrounded coefficients; not an issue for the keys.

## Information-availability
No input violates the rules; values appearing in several places (Table 2 vs Table 3 vs text; Table 6 vs A19) are taken from the table named in each key.