Causal Inference using Python and Machine Learning

専修大学経済学部　英語による特殊講義　前期・後期

Causal Inference using Python and Machine Learning １＆２

This course helps students understand applied causal inference and develop the skills needed to plan and execute their own empirical projects in economics. A key challenge of empirical research in a data-rich environment is formulating insightful questions and conducting robust empirical analysis. To address this, we will examine numerous examples and engage in substantial data analysis ourselves. Topics of the first-semester course include randomized controlled trials, regression and matching techniques. Topics of the second-semester course include instrumental variables, difference-in-differences, and regression discontinuity designs. Throughout the course, we will also explore key contributions from the emerging econometric literature that integrates machine learning with causal inference—referred to as causal machine learning.

By the end of the course, you will have gained a practical familiarity with the tools of causal machine learning, proficiency in data handling using Python, and a solid understanding of the models and methods of applied causal inference.

Topics:

1. Identification and potential outcome framework
2. Review of statistics
3. Python
4. Markdown
5. Analysis and interpretation of randomized trials
6. Regression basics – conditional expectation function
7. Regression basics – casual reg vs causal reg
8. Regression basics – identification
9. Regression basics – estimation
10. Using multivariate regression – omitted variable bias
11. Using multivariate regression – selection on observables
12. Heteroskedasticity and clustered standard errors
13. Directed acyclic graphs (DAG) and variable selection
14. Matchmaker  
15. Inverse probability weighting and Doubly-robust estimator
16. Instrumental variables – selection on unobservables
17. Instrumental variables – heterogeneous effects
18. Instrumental variables – local average treatment effect
19. Panel data model – fixed effects and time effects
20. Panel data model – interactive effects
21. Difference-in-differences method 、DID with multiple time periods
22. Synthetic control method
23. Regression discontinuity design – Sharp RD and Fuzzy RD
24. Least Absolute Shrinkage and Selection Operator (LASSO)
25. Mostly dangerous big data – Double selection method
26. Trees and causal trees
27. Causal forests
28. Targeting
29. Final Exam (presentations, day 1)
30. Final Exam (presentations, day 2)

Textbooks:

• Mastering ‘Metrics – The Path from Cause to Effect, Angrist and Pischke

References:

• Casual Inference for the Brave and True, Matheus Facure
• Applied Causal Inference Powered by ML and AI, Chernozhukov et al. (2024)

Videos:

• Ceteris Paribus: Public vs. Private University
• Randomized Trials: The Ideal Weapon
• How to Read Economics Research Papers: Randomized Controlled Trials (RCTs)
• Selection Bias, Regression, Matching: Will You Make More Going to a Private University?
• Introduction to Instrumental Variables 
• Difference in differences (DD)
• Regression discontinuity designs (RD)
• Mastering Mostly Harmless Econometrics (Alberto Abadie, Joshua Angrist, and Christopher Walters) – 2020 AEA Continuing Education Webcasts

Grading:

• Spring semester (前期): 3 assignments (75%), attendance and class participation (25%)
• Fall semester (後期): 2 assignments (50%), and a term project presentation (50%)