# Data audit
| file | SHA-256 | rows (excl. header) | cols |
|---|---|---|---|
| data/life_expectancy_panel.csv | b3f2f774ed6797034acf62853286ca37fc02334ba3ce70e745e4f3d5c533c3ac | 439 | 4 |
| data/README.md | e9099d209a7642b2b2151161277222719045c3c1094e9490a5d02faa316353d0 | — | — |

Schema: country_name (str, 21 distinct), iso3 (str, 20 distinct), year (int, 2000–2022), life_expectancy (float, years, 3 dp; range 51.49 NGA 2011 – 84.56 JPN 2020; 1 missing).
Anomalies:
1. TUR appears twice (country_name "Turkey" and "Turkiye"), 46 rows = 23 duplicated (iso3, year) pairs with identical values → assembly error from the two extractions; dedupe to one row per (iso3, year), canonical name "Turkiye".
2. BGD, ETH, IND, NGA have rows only for 2011–2022 (12 rows each) → missing 2000–2010; likely from the narrower extraction, but per rules treated as missing; excluded from balanced set.
3. CHE 2007 has an empty life_expectancy cell → missing; CHE excluded.
No exact-duplicate conflicts, no out-of-range years, no implausible values. No input violates the availability rules.
Year counts: 17 rows/yr 2000–2010 (16 countries + dup TUR), 21 rows/yr 2011–2022.