• For more details of our method, see https://arxiv.org/abs/2205.14613 .

• The experiments and methods are developed with Python 3.8, and we recommend to use this version of python inside your virtual environment. First install the required packages with

# R and glmnet is needed for rpy2 and running some of the crt-variants
conda install -c conda-forge r-glmnet 
conda install R pip 

pip install -r requirements.txt

• For Figure 1 and 2 with the qqplot of the decorrelated test statistics and D0-CRT test statistics, run

python exp_qqplot.py

• For the experiments comparing various methods with varying simulation parameters in Figure 3, run

python exp_simu_varying_parameters.py

If you use this code or our method in your project, please use the following citation

Nguyen, B. T., Thirion, B., & Arlot, S. (2022). A Conditional Randomization
Test for Sparse Logistic Regression in High-Dimension. NeuRIPS 2022. arXiv preprint arXiv:2205.14613.