Data audit (P1)
| file | SHA-256 (provided file) | data rows | cols | date range | missing cpi | dup dates | gaps |
|---|---|---|---|---|---|---|---|
| cpi_vintage_2016-07-15.csv | 281fbcdd6a6eeb3e9cb7c1f20476e6bd1f7b925bf140606fe46538fe83f2c83e | 834 | date,cpi | 1947-01..2016-06 | 0 | 0 | none |
| cpi_vintage_2026-10-28.csv | 8220d5b4d70ca32fc2c7b1975651c515967db8a87053cf725159c63ed274cfba | 153 | date,cpi | 2014-01..2026-09 | 1 (2025-10) | 0 | none |
| cpi_assembled.csv | ff732b201a525d6fb171d4e8b2d470f986863c1547a1928e42bb197c0777d44d | 957 | date,cpi,vintage | 1947-01..2026-10 | 1 (stamped 2025-11) | 0 | 2016-07 missing |
Units: index 1982-84=100, SA, monthly, first-of-month dates. No month-on-month move >2% in either vintage file (no obvious typos). Old vintage pre-~2000 values have 1–3 decimals as published. README SHA-256s are of the raw downloads (before reformatting) and do not match the provided files; recorded, not treated as disqualifying.
Anomalies:
- Vintage overlap 2014-01..2016-06 (30 months): levels revised, new−old in [-0.455, 0.330].
- 2026 vintage: empty cpi for 2025-10 → no level; not imputed.
- Assembled file: (a) overlap months use the 2016 vintage instead of the most recent; (b) 2016-07 dropped and all later values stamped one month late (spurious 2026-10 row); 152 value disagreements with the authoritative splice. → Assembled file NOT used for computation (audit only). Pre-2014 values match the 2016 vintage exactly.
Information-availability: no input uses post-2026-10-28 information. Handling: splice = 2016 vintage for 1947-01..2013-12, 2026 vintage for 2014-01..2026-09.