patt-c

This repository provides data and code for reproducing "Estimating population average treatment effects from experiments with noncompliance".

Please cite the paper if you use this code for academic research:

@article{ottoboni2020estimating,
  title={Estimating population average treatment effects from experiments with noncompliance},
  author={Ottoboni, Kellie N and Poulos, Jason V},
  journal={Journal of Causal Inference},
  volume={8},
  number={1},
  pages={108--130},
  year={2020},
  publisher={De Gruyter}
}

Description of scripts in code/

• package-list.R install required packages
• main.R main run list
  • main.sh shell script for main.R
• simulation.R runs simulations and produces results/simulation_res.Rdata
• simulation-plots.R loads results/simulation_res.Rdata and saves to plots/
• prepare-ohie.R merges OHIE data and saves to data/prepare-ohie.Rdata prepare-NHIS.R merges NHIS data and saves to data/prepare-NHIS.Rdata
• prepare-analysis.R imports NHIS and OHIE datasets and creates outcome vectors and common covariates for the analysis; saves to data/prepare-analysis.Rdata
• analysis.R produces empirical estimates; change run.model to TRUE to train complier and response models; saves to data/analysis.Rdata
  • SuperLearner.R Super learner helper functions
  • complier-mod.R Fit complier model if run.model is TRUE
  • complier-mod-cv.R Cross-validate accuracy of complier model if run.model is TRUE
  • response-mod.R Fit response models if run.model is TRUE
  • wtc.R function for weighted t-test with cluster-bootstrapped SEs
• rct-nrt-compare.R produce estimates for Tables A1 and A2
• estimator-compare-plots.R saves heterogeneous treatment effect estimates to plots/
• placebo-test.R produce estimates for Table A3

Instructions

• Clone a copy of the repository to your working directory with the command

$ git clone https://github.com/jvpoulos/patt-c

• The code uses R version 3.5.2 (2018-12-20). To install this R version on Ubuntu, use the command

$ sudo apt-get install r-base-core=3.5.2-1xenial

• Download and extract pretrained response models to results/ directory:
  • response-mod.tar.xz
  • response-mod-patt.tar.xz
• Open package-list.R in a script editor
  • Verify that all required packages in package-list.Rare installed in your R library
• Open main.R in a script editor
  • Change the file path specified by repo.directory to your working directory
  • Change patient.simulation to TRUE to run simulation
  • Save your changes to main.R
• Make shell file main.sh executable from the Linux/Unix command line:

$ chmod +x main.sh

• Execute the file:

$ ./main.sh > main.txt