causalworkshop

Lifecycle: experimental

Demo – causal inference using grf

causalworkshop provides tools for learning causal inference using machine learning methods, here, causal forests.

Installation

Prerequisites

# install devtools if not already installed
if (!requireNamespace("devtools", quietly = TRUE)) {
  install.packages("devtools")
}

# install causalworkshop from GitHub
devtools::install_github("go-bayes/causalworkshop")

Optionally install margot

# install margot for advanced workflows OPTIONAL
# warning, this package is being refactored: for demonstration purposes in the Workshop
devtools::install_github("go-bayes/margot")

This will install causalworkshop along with all required dependencies for the complete workshop experience.

Verifying Installation

After installation, verify everything works:

# Load the package
library(causalworkshop)

# Check prerequisites
check_workshop_prerequisites()

# Run a quick test
workshop_results <- run_workshop(n = 30000, verbose = FALSE)

Troubleshooting

Common Issues and Solutions:

  1. Package compilation errors: Install Rtools on Windows or Xcode command line tools on macOS
  2. Missing dependencies: Run install.packages(c("grf", "dplyr", "ggplot2")) manually
  3. margot package issues: Install latest version with devtools::install_github("go-bayes/margot")
  4. Permission errors: On macOS/Linux, you may need to install packages to a user library

Getting Help:

Quick Start

library(causalworkshop)

# Copy all workshop scripts to your working directory
get_workshop_scripts()

# See what scripts are available
list_workshop_scripts()

# Work through the scripts in order:
# 01-baseline-adjustment.R    - Foundation concepts
# 02-causal-forest-analysis.R - Core methodology  
# 03-rate-qini-curves.R      - Performance evaluation
# 04-policy-trees.R          - Decision rules
# 05-margot-workflow.R       - Professional analysis
# 06-interpretation.qmd       - Results reporting

Option 2: Run the Complete Workshop Function

library(causalworkshop)

# Check prerequisites
check_workshop_prerequisites()

# Run complete educational pipeline
workshop_results <- run_workshop()

# View key results
workshop_results$baseline_results$results_table
workshop_results$causal_forest_results$heterogeneity_stats

# Display educational plots
workshop_results$selection_bias_plot
workshop_results$heterogeneity_plot
workshop_results$targeting_plots$rate_plot

Individual Workshop Modules

# 1. Simulate realistic data
data <- simulate_religious_data(n = 2000)

# 2. Demonstrate selection bias
baseline_results <- workshop_baseline_adjustment(data)
plot_selection_bias(baseline_results$results_table)

# 3. Estimate heterogeneous treatment effects
cf_results <- workshop_causal_forest(data)
plot_heterogeneity_distribution(cf_results$predictions)

# 4. Evaluate targeting performance
targeting_plots <- plot_rate_qini_curves(cf_results$predictions)
targeting_plots$rate_plot
targeting_plots$qini_plot

Workshop Content

Module 1: Selection Bias and Baseline Adjustment

Learn why naive comparisons fail in observational studies and how baseline adjustment addresses selection bias.

# Demonstrates selection bias
baseline_results <- workshop_baseline_adjustment()
baseline_results$bias_reduction  # Quantifies improvement

Module 2: Causal Forests for Heterogeneous Effects

Consider how causal forests can estimate conditional average treatment effects while maintaining valid statistical inference.

# Estimates conditional average treatment effects τ(x)
cf_results <- workshop_causal_forest()
cf_results$heterogeneity_stats   # Summary of effect variation
cf_results$variable_importance   # Drivers of heterogeneity

Module 3: Targeting and Policy Analysis

Explore how treatment effect heterogeneity can inform policy targeting decisions.

# Evaluates targeting performance
plots <- plot_rate_qini_curves(cf_results$predictions)
plots$efficiency_data  # Targeting efficiency metrics

Implementations follow the standards established by the Generalized Random Forests development team.

Citation

If you use this package in academic work, please cite

citation("causalworkshop")

Software License

This R package is licensed under the MIT License. See LICENSE.md for details.

Educational Content License

The educational content, tutorials, and documentation are licensed under Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0).

Copyright © 2025 Joseph A. Bulbulia

Citation

If you use this package in academic work, please cite:

Bulbulia, J.A. (2025). causalworkshop: Educational Package for Causal Inference 
  with Machine Learning. R package version 0.1.0. 
  https://github.com/go-bayes/causalworkshop

This package was developed for educational purposes as part of the Epic Lab at Victoria University of Wellington.

This site is open source. Improve this page.