# Data audit

File: data/life_expectancy_panel.csv — SHA-256 b3f2f774ed6797034acf62853286ca37fc02334ba3ce70e745e4f3d5c533c3ac (matches bundle.json).

• 439 data rows + header, 4 columns: country_name (str), iso3 (str, all 3 chars), year (int, 2000–2022), life_expectancy (float, years, 3 dp; range 51.49–84.56).
• 20 distinct iso3, 21 distinct country_name (TUR appears as both "Turkey" and "Turkiye").
• Duplicates: TUR has 46 rows = 23 years × 2 (one per name); duplicate values identical → deduplicate on (iso3, year).
• Missing: 1 empty cell, CHE 2007.
• Structurally missing years: BGD, ETH, IND, NGA only have rows for 2011–2022 (12 rows each) — the narrower extraction.
• No whitespace/quoting anomalies; no out-of-range values; largest year-on-year changes (≤4.4 y) occur in 2020–2022 (COVID) and are plausible.
• Availability rules: no input violates them; missing values are NOT imputed. data/README.md is provenance only.