This repository contains the code for reproducing simulation and real data analysis results of the paper "Simultaneous inference for generalized linear models with unmeasured confounders".

• gcate: The main module for GCATE.

• ex1: Simulation with Poisson DGP with sample splitting
  • ex1_generate_data.py: Generate simulated data.
  • ex1_run_gcate.py: Run GCATE.
• ex2: Simulation with Poisson DGP without sample splitting
  • ex2_generate_data.py: Generate simulated data.
  • ex2_run_gcate.py: Run GCATE.
  • ex2_run_glm.py: Run GLM oracle and GLM naive.
  • ex2_run_cate.R: Run CATE.
• ex3: Simulation with Splatter simulator
  • ex3_generate_data.py: Generate simulated data.
  • ex3_run_gcate.py: Run GCATE.
  • ex3_run_glm.py: Run GLM naive.
  • ex3_run_cate.R: Run CATE.
• ex4: Lupus data
  • ex4_preprocess_lupus.py: preprocess the lupus data
  • ex4_run_glm.py: Run GLM on subset and full set of covaraites.
  • ex4_run_gcate.py: Run GCATE on subset of covaraites.
  • ex4_run_gcate_full.py: Run GCATE full set of covaraites.
  • ex4_run_cate.R: Run CATE on subset of covaraites.
  • ex4_run_cate_full.R: Run CATE on full set of covaraites.
  • ex4_GO.R: gene ontology analysis
• ex5: Simulation with varying dimensions
  • ex5_blessing_dim.py: Run GCATE on varying dimensions.

• Plot_simu.ipynb: Reproduce the figures and tables for simulation studies.
• Plot_lupus.ipynb: Reproduce the figures and tables for the lupus data analysis.

The following packages are required for the reproducibility workflow.

For simulation studies, the workflow is as follows:

• Run script ex1_generate_data.py to generate simulated data, which will be stored in the folder /data/ex1/. The data for the second and the third experiments can be similarly generated by running ex2_generate_data.py and ex3_generate_data.py, respectively.
• Run scripts of individual methods for each experiment as described below, and the results will be stored in the folder result/:
  • Ex1: ex1_run_gcate.py
  • Ex2: ex2_run_glm.py, ex2_run_gcate.py, ex2_run_cate.R
  • Ex3: ex3_run_glm.py, ex3_run_gcate.py, ex3_run_cate.R
• For experiments on varying dimensions, run ex5_blessing_dim.py.
• Use Plot_simu.ipynb to reproduce the figures (Figures 2-6, F1-F2, and G3) and table (Table G2) based on the previous results.

For real data analysis, the workflow is as follows:

• Obtain the h5ad file of the lupus data from the authors of the original paper and store it in the folder data/lupus/GSE174188_CLUES1_adjusted.h5ad.
• Run ex4_preprocess_lupus.py to preprocess the lupus data.
• Run scripts of individual methods (ex4_run_glm.py, ex4_run_gcate.py, ex4_run_gcate_full.py, ex4_run_cate.R, ex4_run_cate_full.R), and the results will be stored in the folder result/lupus/.
• Use Plot_lupus.ipynb to reproduce the figures (Figures 6, G4-G9, and G11) and tables (Tables G3-G4) based on the previous results.
• Run ex4_GO.R to perform gene ontology analysis (Figure G10).