https://github.com/cyberagentailab/dte-ml-adjustment
Code for "Estimating Distributional Treatment Effects in Randomized Experiments: Machine Learning for Variance Reduction"
https://github.com/cyberagentailab/dte-ml-adjustment
Last synced: 9 months ago
JSON representation
Code for "Estimating Distributional Treatment Effects in Randomized Experiments: Machine Learning for Variance Reduction"
- Host: GitHub
- URL: https://github.com/cyberagentailab/dte-ml-adjustment
- Owner: CyberAgentAILab
- License: mit
- Created: 2024-05-24T04:12:10.000Z (about 2 years ago)
- Default Branch: main
- Last Pushed: 2024-05-27T05:15:50.000Z (about 2 years ago)
- Last Synced: 2025-09-10T07:42:51.564Z (9 months ago)
- Language: R
- Homepage:
- Size: 4.99 MB
- Stars: 2
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
## Estimating Distributional Treatment Effects in Randomized Experiments: Machine Learning for Variance Reduction
This repository contains code to replicate the experimental results from "Estimating Distributional Treatment Effects in Randomized Experiments: Machine Learning for Variance Reduction."
### Folders
1. `data` folder includes files to create dataset used for empirical application from [Ferraro & Price (2013)](https://direct.mit.edu/rest/article-abstract/95/1/64/58053/Using-Nonpecuniary-Strategies-to-Influence). Download original data from https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN1/22633&version=1.1 and save `090113_TotWatDat_cor_merge_Price.dta` file in data folder.
2. `experiment` folder contains all R files used for analysis
### Experiment Files
1. `functions.R` file includes all necessary functions
2. `run_simulation.R` includes code to run the Monte Carlo simulations and saves results as .rds files
3. `compute_stats.R` includes code to calculate evaluation metrics (e.g. bias, RMSE) from the saved simulation results (.rds files) and saves them as .csv files
4. `plot_figures.R` includes code to load the .csv files and plot figures for the simulation study
3. `experiment_water_consumption.R` includes code to replicate the analysis of experimental data from Ferraro & Price (2013)
### Instructions
1. Install all necessary packages in R
2. To replicate the results from the
Monte Carlo simulation, run the files in the following order: (1) `run_simulation.R`, (2) `compute_stats.R`, (3) `plot_figures.R`. The outputs will be figures appeared in Figures 1, 3 and 4 in the paper.
3. Run `experiment_water_consumption.R` to replicate the results from the water consumption experiment. The output will be figures appeared in
Figure 2 in the paper.
### R version and attached packages
- R version 4.3.1
- `RColorBrewer_1.1-3` `ggpubr_0.6.0` `fastglm_0.0.3` `bigmemory_4.6.1` `xgboost_1.7.5.1` `foreign_0.8-84` `ggplot2_3.4.3` `dplyr_1.1.2` `doParallel_1.0.17` `glmnet_4.1-8` `Matrix_1.6-1.1` `doMC_1.3.8` `iterators_1.0.14` `foreach_1.5.2` `grf_2.3.1` `randomForest_4.7-1.1` `gridExtra_2.3` `tidyr_1.3.0` `haven_2.5.3` `readr_2.1.4`