User Guide

How to choose among the methods CausalML implements and the theory behind them, how to interpret and validate what a fitted model produces and with what uncertainty, the benchmark datasets with known ground truth, and worked examples as runnable notebooks.

• Choosing an Estimator
  • Start from the data
  • Capability matrix
  • Two rules of thumb
• Methodology
  • Supported Algorithms
  • Meta-Learner Algorithms
  • CATE Model Evaluation
  • Tree-Based Algorithms
  • Neural Network Algorithms
  • Value optimization methods
  • Probabilities of causation
  • Selected traditional methods
  • Targeted maximum likelihood estimation (TMLE) for ATE
• Interpretable Causal ML
  • Meta-Learner Feature Importances
  • Uplift Tree Visualization
  • Uplift Tree Feature Importances
• Validation
  • Checking Overlap
  • Validation with Uplift Curve (AUUC)
  • Testing for Heterogeneity with RATE
  • Model Selection with Validation Losses
  • Validation with Ground Truth
  • Validation with Synthetic Data Sets
  • Validation with Multiple Estimates
  • Validation with Sensitivity Analysis
• Uncertainty Quantification
  • Average treatment effects
  • CATE estimates
  • Evaluation metrics
  • What an interval does and does not cover
• Benchmark Datasets
  • Which metric applies to which dataset
  • Datasets with a loader
  • Datasets without a loader
• Examples
  • Getting started
  • Meta-learners
  • Uplift and causal trees
  • Evaluation and validation
  • Interpretation and feature selection
  • Neural models
  • Instrumental variables
  • Policy and value optimization
  • Datasets and benchmarks
• References
  • Open Source Software Projects
  • Papers