Simulation dataset on power and size of Wald, likelihood-ratio, and score test statistics in parametric competing risks models under hybrid censoring
Description
This dataset contains Monte Carlo simulation results for parametric competing risks models (CRM) with K=2 or K=3 independent competing causes of failure under hybrid (Type I/II combined) censoring. The dataset was generated as part of the doctoral research at the National University of Uzbekistan. The archive includes 13 CSV files covering: - Bias and RMSE of the maximum likelihood estimator (MLE) for exponential, Weibull (shape alpha=2), and Gompertz families; - Empirical Type I error and power of Wald (Wn), likelihood-ratio (LRn), and score (Sn) test statistics at nominal level alpha=0.05; - Non-centrality parameter lambda* and theoretical power under Pitman local alternatives; - Coverage and mean length of asymptotic and parametric bootstrap 95% confidence intervals; - Robustness to risk dependence via Clayton copula (theta_cop in {0, 0.3, 0.5, 1.0}); - Minimum required sample size n* for a given precision target; - Application to the publicly available lung cancer dataset (survival::lung, n=228). Experimental grid: n in {50, 100, 200, 500}, beta in {0.25, 0.50, 0.75, 1.00}, R=5000 replications per setting. Python 3.11, NumPy 1.24, SciPy 1.11; R 4.3.2 (survival, cmprsk). Base random seed 42 with independent SeedSequence streams per parallel worker.
Files
Steps to reproduce
1. Download all files from this repository. 2. Install required Python packages: pip install numpy==1.24 scipy==1.11 joblib==1.3 pandas==2.0 matplotlib==3.8 3. Run the main simulation script: python simulate_crm.py --seed 42 --R 5000 --n_jobs -1 This generates all CSV tables in ./data/ (approx. 4 hours on 8-core machine). 4. Run bootstrap CI script: python bootstrap_ci.py --R_boot 999 --R_sim 2000 This generates table_ci_coverage.csv (approx. 2 additional hours). 5. For the lung cancer analysis, run in R: Rscript lung_analysis.R Ensure R packages survival and cmprsk are installed. 6. All outputs will appear in the ./data/ directory. 7. For full documentation, see README.md.
Institutions
- National University of UzbekistanTashkent, Tashkent