Data audit (P1)
| file | SHA-256 (computed) | matches README | data rows | cols | date range | missing | dup dates | gaps |
|---|---|---|---|---|---|---|---|---|
| data/UNRATE.csv | 621ff6fadf50910cb7a21ee20ba610c247cb043b29b4e05c4c1a72fadca2a06c | yes | 945 | 2 (observation_date, UNRATE) | 1948-01-01 – 2026-09-01 monthly | 1 (2025-10-01 empty field) | 0 | none (complete monthly calendar, monotone) |
| data/GDPC1.csv | 1c6c6e9f59e383e20c849dd299175c7a307526c09abdc18f18da9754fc5b3343 | yes | 319 | 2 (observation_date, GDPC1) | 1947Q1 – 2026Q3 quarterly | 0 | 0 | none |
Units: UNRATE percent (SA); GDPC1 billions of chained 2017 USD, SAAR. Ranges: UNRATE 2.5–14.8; GDPC1 2172.4–24611.1. No non-numeric values except the one empty field.
Anomaly: UNRATE 2025-10 is empty (outside the 1959Q4–2019Q4 level window) → no effect; recorded in results data_issues.
Information-availability: files contain data after 2019Q4 and before 1959Q4; these are excluded by exclusion step 1 (only 1959Q4 levels used as the differencing base). No file violates the rules once filtered.
Within the window 1959Q4–2019Q4 every quarter has 3 monthly UNRATE values and a GDPC1 value (asserted in analysis.py).
Exclusion ledger unit: quarters in the merged quarterly panel (union of UNRATE quarters 1948Q1–2026Q3 and GDPC1 quarters 1947Q1–2026Q3 = 319). 319 − 78 − 1 = 240 = n_analysed.