# Data audit (P1)
- File: data/UNRATE.csv, SHA-256 621ff6fadf50910cb7a21ee20ba610c247cb043b29b4e05c4c1a72fadca2a06c (matches README).
- 945 data rows + header, 2 columns: observation_date (YYYY-MM-01 string), UNRATE (float, percent, SA).
- Range 1948-01-01 .. 2026-09-01, monthly, no gaps in dates, no duplicate dates, all dates on day 01.
- Missing: 1 value, 2025-10-01 (empty field). Affects: naive/model at origin 2025-10 (use latest earlier obs 2025-09); actual for (origin 2025-09,h1) and (origin 2025-07,h3) -> excluded from error stats; training windows for origins >= 2025-10 (regression rows involving the missing value dropped).
- Values: min 2.5, max 14.8 (2020-04, COVID spike: 3.5 -> 4.4 -> 14.8 -> 13.2). Genuine data, kept; it is in all training windows. Noted as anomaly; no outlier treatment (decided a priori, not tuned on evaluation errors).
- Information availability: file is final vintage retrieved 2026-10-30, so all values are revised; violates real-time availability in the vintage sense. Handling: as TASK instructs, evaluate on given data and make no real-time claims. Observations after each origin are masked in code.