Checks
- Rerun of analysis.py → results.json byte-identical (cmp): PASS.
- Independent recomputation via pivot_table/dropna: 15 balanced countries, same list, means 78.060267/81.553067, change 3.4928, baseline 78.692 over 20 countries: PASS.
- analytic.csv: 345 rows = 15×23, no duplicate (iso3, year), each country 23 rows: PASS.
- Exclusion ledger: 439→416→416→415→345, n_before−n_removed=n_after each row, chains: PASS (asserted in script).
- n_countries_balanced + n_countries_incomplete = 15+5 = 20 = n_countries_total: PASS.
- change = mean_2022 − mean_2000 (3.4928): PASS.
- Magnitude vs baseline: balanced 2022 mean (81.55) > all-available (78.69) as expected since excluded set contains 4 lower-LE countries (BGD, ETH, IND, NGA): PASS (plausible).
- Duplicate conflict check: TUR duplicates have identical values (script aborts otherwise): PASS.