Data audit (P1)
| file | SHA-256 | rows (excl. header) | cols | range |
|---|---|---|---|---|
| data/cpi_quarterly.csv | c8652dec1567ec35e3b84ee8fa55545e9bf0ae4990a4d6d05b89a6d22f6801ea (matches README) | 127 | 3 (quarter date, cpi_avg float index 2002=100, infl_annualised float % a.r.) | 1995-01-01 … 2026-07-01 |
| data/cpi_monthly.csv | 855d069176369d30d02cbbe0cc6456b76d2f3967cdaabc1b362c2461ef9015a2 (matches README) | 381 | 2 (month date, cpi float, 1 decimal) | 1995-01-01 … 2026-09-01 |
- Missing: only infl_annualised for 1995Q1 (by construction, no prior quarter). No missing in monthly file.
- Duplicates: none in either file. Spacing: strictly consecutive quarters / months, no gaps.
- Every quarter has 3 months; recomputed quarterly means from the monthly file equal cpi_avg exactly (rounded to 6 dp).
- Recomputed 400·Δln(cpi_avg) matches infl_annualised to within 3.3e-6 (rounding of cpi_avg to 6 dp); the file values are used as given.
- infl_annualised: n=126, mean 2.13, sd 2.39, min −6.04 (2008Q4), max 12.15 (2022Q2). Extremes are genuine episodes (GFC oil-price collapse, 2022 inflation surge), not errors; kept.
- Strong seasonal pattern expected (NSA index).
- Information availability: files contain data beyond the last origin (through 2026Q3); these are needed only as targets (latest target 2025Q3+4 = 2026Q3). The analysis code slices the series to quarters ≤ origin before any estimation; this is enforced programmatically and checked in P4.
- Exclusion: 1995Q1 (no inflation value) removed from the inflation series → 126 usable quarters.