# Data audit
- `data/life_expectancy_panel.csv` SHA-256 b3f2f774ed6797034acf62853286ca37fc02334ba3ce70e745e4f3d5c533c3ac; 439 data rows, 4 columns (country_name str, iso3 str, year int, life_expectancy float, years). Years 2000–2022. 20 distinct iso3 codes, 21 distinct names.
- `data/README.md` SHA-256 e9099d209a7642b2b2151161277222719045c3c1094e9490a5d02faa316353d0 (provenance only).
- Value range 51.49–84.56 years (plausible). Year-over-year jumps over 1.5 years occur only around 2020–2022 (COVID; e.g. IND, USA, ESP), so they are plausible and kept.
- Anomaly 1: TUR has 46 rows, each year appearing as both "Turkey" and "Turkiye" with identical values. This is an assembly duplicate. Deduplicated on (iso3, year), keeping the last row ("Turkiye").
- Anomaly 2: CHE 2007 has an empty life_expectancy cell → missing, not imputed. CHE is therefore not balanced.
- Anomaly 3: BGD, ETH, IND and NGA have rows only for 2011–2022 (likely the narrower extraction) → not balanced.
- No other duplicates, no whitespace in codes, no out-of-range years.
- Information-availability: no input violates the rules. Only the CSV is used.