Reproducibility Materials for "Improved Risk Estimation Need Not Improve Finite-Budget Selection"

Reproducibility materials for the manuscript “Improved Risk Estimation Need Not Improve Finite-Budget Selection: An Analytically Anchored Nested-CVaR Example”, by Haowei Zeng. This synthetic diagnostic example studies selection between two measurement designs using the 90% conditional value-at-risk (CVaR) of expected interval scores across 24 equally weighted scenarios. The estimand is CVaR over scenario expectations, not CVaR over individual scores. In a fixed configuration with a 1% true risk margin, increasing the inner simulation budget reduced both designs’ risk-estimation bias and RMSE while increasing wrong-design selection frequencies over the three tested budgets. The materials include analytically anchored reference calculations, archived simulation results, structural controls, and scripts for reproducing the core figures. Files and licenses:• code_MIT.zip — Python code licensed under the MIT License. Third-party rights are excluded.• data_figures_CC_BY_4_0.zip — synthetic data, configuration and original figures licensed under CC BY 4.0, to the extent the author holds rights to license them. Extract both archives into the same working directory, preserving their folder structures and separate license notices. Required Python packages are NumPy, pandas and Pillow. Running draw_core_figures_v32.py regenerates the core figures from the archived data without running new simulations. This is a limited diagnostic case, not a new selection algorithm or a general claim that larger budgets worsen decisions. It does not establish clinical or real-world validity. All data are synthetic; no patient data are included. The associated manuscript has not undergone peer review. Generative AI substantially assisted formulation, derivation, coding, literature inspection and writing. AI checks are not external peer review, and AI is not an author. The work was self-funded by the author. No external funding was received. The author declares no competing interests.