Simulation and application data: Bayesian and maximum likelihood estimation of the Gamma distribution under random right censoring

Published: 16 October 2026| Version 1 | DOI: 10.17632/hsmjmrfn8t.1
Contributors:
,
,
,
,

Description

Replication-level output, performance summaries and validation results of two Monte Carlo experiments comparing maximum likelihood (EM algorithm) and Bayesian (posterior quadrature) estimation of the Gamma shape and rate under Type I, random, informative and dependent right censoring (108,000 simulated data sets; 480,000 estimate rows), together with the Channing House data for 97 male residents (left-truncated and right-censored) and the output of their analysis, including 480,000 posterior draws. The complete Python code, a verification script, a worked example and a data dictionary (284 columns) are included. Start with README.md.

Files

Steps to reproduce

Unzip the archive and read README.md. Install Python 3.12 and the packages in code/requirements.txt (pip install -r requirements.txt). From the code/ folder: (1) python verify_quick.py (about 1 min) regenerates 1,275 replication-level rows from their random-number stream keys and checks them against data/; (2) python reproduce_tables.py recreates all tables and the figure of the data article from data/; (3) python example_usage.py follows one simulation setting through the files. The full simulation is rerun with run_simulation.py, run_validation.py, run_validation_long.py, run_benchmark.py and run_real_data.py, followed by make_dataset.py. Column definitions are in data_dictionary.csv. Checksums: sha256sum -c checksums_sha256.txt.

Categories

Statistics, Econometrics, Bayesian Statistics, Survival Analysis, Application of Monte Carlo Method

Licence