Machine Learning-based Causal Inference Tutorial

[chengjun]

https://bookdown.org/stanfordgsbsilab/ml-ci-tutorial/

This tutorial will introduce key concepts in machine learning-based causal inference. It’s an ongoing project and new chapters will be uploaded as we finish them. Topics currently covered:

1. Introduction to Machine Learning
2. Average Treatment Effects
3. Heterogeneous Treatment Effects
4. Policy Evaluation
5. Policy Learning
6. Causal Panel Data
7. Matrix Completion Methods

Machine Learning & Causal Inference: A Short Course

This course by Susan Athey, Jann Spiess and Stefan Wager is the companion course to this tutorial.

The course includes a series of videos and slides that cover a quarter-long course at Stanford. The course is designed for students and researchers looking to learn more about how machine learning can be used to measure the effects of interventions, understand the heterogeneous impact of interventions, and design targeted treatment assignment policies.

1 Introduction

• 1.1 Getting started
• 1.2 Changelog
• 1.3 Acknowledgments

2 Introduction to Machine Learning

• 2.1 Key concepts
• 2.2 Common machine learning algorithms
• 2.3 Further reading

3 ATE I: Binary treatment

• 3.1 Notation and definitions
• 3.2 Difference-in-means estimator
• 3.3 Turning an experimental dataset into an observational one
• 3.4 Direct estimation
• 3.5 Inverse propensity-weighted estimator
• 3.6 Augmented inverse propensity-weighted (AIPW) estimator
• 3.7 Diagnostics
• 3.8 Further reading

4 HTE I: Binary treatment

• 4.1 Pre-specified hypotheses
• 4.2 Data-driven hypotheses
• 4.3 Assessing heterogeneity with RATE
• 4.4 Further reading

5 Policy Evaluation I - Binary Treatment

• 5.1 Via sample means
• 5.2 Via AIPW scores
• 5.3 Case study
• 5.4 Further reading

6 Policy Learning I - Binary Treatment

• 6.1 Non-parametric policies
• 6.2 Parametric policies
• 6.3 Case study
• 6.4 Topics 1: Subgroups using learned policy
• 6.5 Topics 2: Learning with uncertain costs
• 6.6 Further reading

7 Causal Panel Data

• 7.1 Data Setup
• 7.2 Alternative methods
• 7.3 Further reading

8 Matrix Completion Methods

• 8.1 Set Up
• 8.2 Patterns of missing data
• 8.3 Thin and Fat Matrices
• 8.4 Horizontal and Vertical Regressions
• 8.5 Fixed Effects and Factor models
• 8.6 Machine Learning on Matrix Completion
• 8.7 Further Reading

9 Additional Resources

• 9.1 Machine Learning & Causal Inference: An Introductory Course
• 9.2 Data from Experiments
• 9.3 Applications of the Methods to Behavioral Science and Social Impact
• 9.4 Useful software packages
• 9.5 Bandit Experiment Simulation Tool