Videos and books for causal inference

Every resource on this page is free to watch or read, from the author's, the university's or the publisher's own page, unless it is marked "(book)". Each lesson of the track also ends with a Watch and read section that names the exact video, chapter or paper.

How to use it. Watch a short video for the picture, then read the lesson and run its notebook. After that, read the matching chapter from the table at the end of this page. Brady Neal's course follows almost the same order as this track, so it is the best single companion. What If (Hernán and Robins) is the best book for the potential-outcome side; Causal Inference in Statistics: A Primer (Pearl, Glymour and Jewell) is the best short book for the graph side.

A note on notation. This track writes the treatment as $T$, the outcome as $Y$, covariates as $X$ and potential outcomes as $Y(1), Y(0)$. Books use other letters for the same objects: Hernán and Robins write $A$ for the treatment and $Y^{a}$ for a potential outcome; economists (Angrist and Pischke, Cunningham) write $D$; Imbens and Rubin write $W$; Pearl often writes $X$ for the treatment and $Y_x$ for a potential outcome. The ideas are the same. Lesson 02 gives the full list.

Video courses #

Start here #

• Brady Neal, Introduction to Causal Inference (course page, full lectures playlist, short-video playlist). A full machine-learning-oriented course with free course notes: potential outcomes, graphs, the do-operator, estimation, instruments, difference-in-differences, causal discovery, transfer learning and counterfactuals. Guest talks by Susan Athey (heterogeneous effects) and Alberto Abadie (synthetic control). It pairs with almost every lesson; see the table below.
• Mod•U, Causal Inference Bootcamp (playlist). Videos of 3–9 minutes from Duke University, without formulas: correlation and causation, randomized experiments, matching, instruments, noncompliance, regression discontinuity and counterfactuals. Watch one before a lesson when its topic is new to you.

Graphs, confounders and controls #

• Richard McElreath, Statistical Rethinking 2023 (playlist, course materials). A Bayesian statistics course that draws a causal graph before every model. Lectures 5 and 6 are the causal part and pair with lessons 03 and 04; the rest of the course is Bayesian modeling.

Causal inference for machine learning #

• MIT 6.S897, Machine Learning for Healthcare (2019) (course page, YouTube playlist). Lectures 14 and 15 are the two videos above. The rest of the course shows where causal questions appear in clinical machine learning.
• Stanford GSB, Machine Learning & Causal Inference: A Short Course (course page, YouTube playlist). Susan Athey and Stefan Wager on the ATE with ML, heterogeneous effects, causal forests and policy learning, with R tutorials. It pairs with lessons 05 and 08.

Designs from economics: instruments, cutoffs and panels #

• Marginal Revolution University, Mastering Econometrics with Joshua Angrist (playlist, course page). Short, clear videos on randomized trials, regression, instrumental variables, regression discontinuity and difference-in-differences. They pair with lessons 02, 06 and 07.
• Nick Huntington-Klein, The Effect videos (playlist). One 10-minute video per chapter of his free book The Effect: causal diagrams, matching, instruments, regression discontinuity, difference-in-differences and event studies. They pair with lessons 03, 05, 06 and 07.

Free books #

• Miguel Hernán and James Robins, Causal Inference: What If (book page, free PDF). The standard text for potential outcomes in epidemiology and medicine. Part I (chapters 1–10) needs no models and is very clear; Part II covers IPW, standardization, the g-formula and instruments. Pairs with lessons 01, 02, 03 and 05.
• Brady Neal, Introduction to Causal Inference from a Machine Learning Perspective (course notes, free PDF). Short course notes that follow the video course: potential outcomes, graphs, do-calculus, estimation, sensitivity analysis, instruments, difference-in-differences, causal discovery and counterfactuals.
• Jonas Peters, Dominik Janzing and Bernhard Schölkopf, Elements of Causal Inference (MIT Press, 2017; open access, free PDF from MIT Press and OAPEN). Structural causal models, cause-effect models with two variables, causal discovery, and the links to machine learning (causal and anticausal learning, semi-supervised learning, covariate shift). Pairs with lessons 09, 10 and 11.
• Scott Cunningham, Causal Inference: The Mixtape (free online). An economics view with code in R, Stata and Python: potential outcomes and randomization, graphs, matching, regression discontinuity, instruments, difference-in-differences and synthetic control. Pairs with lessons 02, 05, 06 and 07.
• Nick Huntington-Klein, The Effect: An Introduction to Research Design and Causality (free online). Very readable. Part I teaches how to draw and use causal diagrams; Part II covers matching, fixed effects, event studies, difference-in-differences, instruments and regression discontinuity.
• Matheus Facure, Causal Inference for the Brave and True (free online). Python notebooks for every method, with a second part on heterogeneous effects, meta-learners and double ML for industry problems. Pairs with lessons 05 and 08.
• Victor Chernozhukov, Christian Hansen, Nathan Kallus, Martin Spindler and Vasilis Syrgkanis, Applied Causal Inference Powered by ML and AI (free online). Double ML, orthogonal scores, heterogeneous effects and graphs, with Python and R notebooks. The main reference for lesson 08.
• Stefan Wager, Causal Inference: A Statistical Learning Approach (free PDF). Lecture notes from Stanford: randomized trials, doubly robust estimation, heterogeneous effects, the R-learner, policy learning, instruments and panel data, with proofs. Pairs with lessons 05, 06 and 08.

Books to know (not free) #

• Judea Pearl, Madelyn Glymour and Nicholas Jewell, Causal Inference in Statistics: A Primer (2016, book; companion page with free solutions). About 130 pages: graphs, the do-operator, backdoor and frontdoor adjustment, counterfactuals. Pairs with lessons 03, 04 and 09.
• Judea Pearl, Causality: Models, Reasoning, and Inference (2nd ed. 2009, book). The full theory of structural causal models, do-calculus and counterfactuals.
• Guido Imbens and Donald Rubin, Causal Inference for Statistics, Social, and Biomedical Sciences (2015, book). Potential outcomes, randomization inference and matching in depth.
• Joshua Angrist and Jörn-Steffen Pischke, Mostly Harmless Econometrics (2009, book). Regression, instruments, difference-in-differences and regression discontinuity, from the economics side.
• Ron Kohavi, Diane Tang and Ya Xu, Trustworthy Online Controlled Experiments (2020, book; book page). A/B tests in industry: metrics, sample-ratio mismatch, interference and common pitfalls. Pairs with lesson 02.
• Judea Pearl and Dana Mackenzie, The Book of Why (2018, book). A popular introduction without formulas; the source of the "ladder of causation" in lesson 01.

Short reads #

• Carlos Cinelli, Andrew Forney and Judea Pearl (2022), "A Crash Course in Good and Bad Controls". Many small graphs that show which variables to adjust for and which to avoid. Read it after lesson 04.
• Paul Holland (1986), "Statistics and Causal Inference", Journal of the American Statistical Association. The paper that named the fundamental problem of causal inference (lesson 02).
• Chernozhukov et al. (2018), "Double/debiased machine learning for treatment and structural parameters". The double ML paper (lesson 08).
• Künzel, Sekhon, Bickel and Yu (2019), "Meta-learners for estimating heterogeneous treatment effects using machine learning". The S-, T- and X-learners (lesson 08).
• Wager and Athey (2018), "Estimation and inference of heterogeneous treatment effects using random forests". Causal forests (lesson 08).
• Zheng, Aragam, Ravikumar and Xing (2018), "DAGs with NO TEARS: Continuous optimization for structure learning". NOTEARS (lesson 10).
• Schölkopf et al. (2012), "On causal and anticausal learning". Why the direction of prediction matters for shift and semi-supervised learning (lesson 11).
• Schölkopf et al. (2021), "Toward causal representation learning". A survey of causality and machine learning (lesson 11).

Which lesson pairs with what #