We provide code to reproduce all the experiments presented in "Cross-Study Replicability in Cluster Analysis" Statist. Sci. 38(2): 303-316 (May 2023). DOI: 10.1214/22-STS871. See also https://arxiv.org/pdf/2202.01910.pdf).
The repository is divided into 2 main folders:
synthetic_data/which contains all the python scripts and data to replicate the analysis on synthetic data.real_data/which contains all the python scripts and data to replicate the analysis on real cancer data.
.
|____synthetic_data
| |____Appendix_Additional_Experiments.ipynb
| |____Appendix_Competing_Prediction_Accuracy.ipynb
| |____Appendix_Competing_Stability.ipynb
| |____Appendix_Competing_Tests.ipynb
| |____Plots.ipynb
| |____Fit.ipynb
| |____utils_synthetic.py
| |____Results
| |____Data
| | |____Gaussians
| | |____Shapes
| |____Plots
|____real_data
| |____Data_Processing.ipynb
| |____Fit.ipynb
| |____Plots.ipynb
| |____utils_real_data.py
| |____Data_Processing.R
| |____Data
| |____Results
The synthetic data used can be found in synthetic_data/Data (Gaussians contains high dimensional Gaussian data, while Shapes contains the benchmark datasets from P. Fränti and S. Sieranoja, "K-means properties on six clustering benchmark datasets", Apllied Intelligence 2018).
The real data used can be found in synthetic_data/r_data, where the Mainz, Transbig and VDX cancer data can be found.
In both synthetic_data/ and real_data/ the iPython notebook Fit.ipynb contains all the code needed in order to fit the experiments and save the data necessary to then reproduce the plots in the main text.
In both synthetic_data/ and real_data/ the iPython notebook Plots.ipynb contains all the code needed in order to reproduce the plots in the main text. The intermediary files accessed are stored in the Results/ folder.
To replicate the experiments in the Appendix, run the synthetic_data/Appendix_* iPython notebooks. These notebooks contain both code to run (i) generate synthetic data, (ii) run the analyses and (iii) reproduce plots.