Replication package for: Screening short annual regional statistics: finite-sample calibration and comparison of robust outlier rules, with a case study from Uzbekistan
Annotatsiya
Replication package for the article "Screening short annual regional statistics: finite-sample calibration and comparison of robust outlier rules, with a case study from Uzbekistan". The package contains Python code that tabulates finite-sample critical values for the median–MAD and raw median–Qn screening rules (calibrate.py), compares them with the t prediction rule under heavy tails, autocorrelation and contaminated histories (compare.py), and evaluates alerts and injected breaks on regional tables of Uzbekistan (evaluate.py). It also includes a reproducible release pipeline (rsp.py: ingest, validate, screen, steward review, compute, publish) with a SHA-256 manifest, the steward decision file, chain-link sensitivity variants for the 2023 break (chainlink.py), and the code for all figures. Source tables (raw/) are per-capita regional series computed from data published by the National Statistics Committee of the Republic of Uzbekistan (stat.uz), retrieved in August 2026. See README.md for run instructions.
Hali tarjima qilinmagan