Doubly Robust Conditional Independence Testing with Generative Neural Networks

This repository contains the code and implementation details for the manuscript
"Doubly Robust Conditional Independence Testing with Generative Neural Networks."

---

Repository Overview

The project is organized into two main parts:

1. Simulations/

Contains code for the following sections of the paper:
Section_4_1, Section_4_2, Section_4_3, Appendix_B, Appendix_C, Appendix_D, and Appendix_E.

2. Real_Data_Applications/

Contains code for the following sections of the paper:
Section_5_1, Section_5_2, and Appendix_F.

Each folder includes code and a corresponding README.md file that provides detailed instructions on how to reproduce the figures or tables presented in the paper.

---

Data Availability

• For Section_5_1 and Section_5_2, the MNIST dataset (.pt files) is provided in the MNIST/ folder under Section_5_1.

• For Appendix_F, the CCLE dataset can be obtained from the GCIT repository. Please download the following datasets:

  • Response data
  • Mutation data
  • Expression data

Place these datasets in the appropriate locations as described in the relevant README.md files.

---

Environment and Dependencies

Python Environment

The following package versions were used:

Package	Version
python	3.10.x
torch	2.0.1+cu118
numpy	~1.25
scipy	~1.10.1
tqdm	~4.65
scikit-learn	1.3.2
tensorflow_probability	~0.21.0

R Environment

The following R version was used for all R Markdown (.Rmd) or R script files:

• R version 4.3.3 (2024-02-29 ucrt)