Introduction to Causal Inference

You’ve found the online causal inference course page. Although, the course text is written from a machine learning perspective, this course is meant to be for anyone with the necessary prerequisites who is interested in learning the basics of causality. I do my best to integrate insights from the many different fields that utilize causal inference such as epidemiology, economics, political science, machine learning, etc. You can see the tentative course schedule below.

You can join the course Slack workspace where you can easily start discussions with other people who are interested in causal inference. For information about office hours, see the office hours section below. If you’re interested in leading a reading group discussion, check out the suggested reading group papers to see if one piques your interest. When emailing me about this course, please include “[Causal Course]” at the beginning of your email subject to help make sure I see your email. If you want to receive course updates, sign up for the course mailing list. The main textbook we’ll use for this course is Introduction to Causal Inference (ICI), which is a book draft that I’ll continually update throughout this course.

Course Schedule (tentative)

Note about slides: they currently don’t work well with Adobe Acrobat, though they seem to work with other PDF viewers.

Week	Topics	Lecture	Readings	Reading Group Paper
August 31	Motivation Course Preview Course Information	Video Slides Info	Chapter 1 of ICI	None
September 7	Potential Outcomes A Complete Example with Estimation	Video Slides	Chapter 2 of ICI	Does obesity shorten life? The importance of well-defined interventions to answer causal questions (Hernán & Taubman, 2008)
September 14	Graphical Models	Video Slides	Chapter 3 of ICI	Does Obesity Shorten Life? Or is it the Soda? On Non-manipulable Causes (Pearl, 2018)
September 21	Backdoor Adjustment Structural Causal Models	Video Slides	Chapter 4 of ICI	Single World Intervention Graphs: A Primer (Richardson & Robins, 2013)
September 28	Randomized Experiments Frontdoor Adjustment do-calculus Graph-Based Identification	Video Slides	Chapters 5-6 of ICI	On Pearl’s Hierarchy and the Foundations of Causal Inference (Bareinboim et al., 2020)
October 5	Estimation Susan Athey Guest Talk - Estimating Heterogeneous Treatment Effects (Oct 8th at 3 - 4 pm EDT)	Video Slides Guest Talk	Chapter 7 of ICI	Adapting Neural Networks for the Estimation of Treatment Effects (Shi, Blei, Veitch, 2019)
October 12	Unobserved Confounding, Bounds, and Sensitivity Analysis	Video Slides	Chapter 8 of ICI	Sense and Sensitivity Analysis: Simple Post-Hoc Analysis of Bias Due to Unobserved Confounding (Veitch & Zaveri, 2020)
October 19	Instrumental Variables	Video Slides	Chapter 9 of ICI	Deep IV: A Flexible Approach for Counterfactual Prediction (Hartford et al., 2017)
October 26	Difference-in-Differences Alberto Abadie Guest Talk - Synthetic Control (Oct 29th at 10 - 11 am EDT)	Video Slides Guest Talk	Chapter 10 of ICI	Regression Discontinuity Designs in Economics (Lee & Lemieux, 2010)
November 2	--- Break Week - No Lecture ---	None	Past Readings	None
November 9	Causal Discovery from Observational Data Jonas Peters Guest Talk (November 13 at 10 am EST)	Video Slides	Chapter 11 of ICI	Inferring causation from time series in Earth system sciences (Runge et al., 2019)
November 16	Causal Discovery from Interventions	Video Slides	Chapter 12 of ICI (Coming soon)	Permutation-based Causal Inference Algorithms with Interventions (Wang et al., 2017)
November 23	Transfer Learning Transportability	Video Slides	Chapter 13 of ICI (Coming soon)	A causal framework for distribution generalization (Christiansen et al., 2020)
November 30	Yoshua Bengio Guest Talk - Causal Representation Learning (Dec 1st at 1 - 2:30 pm EST)	Guest Talk Slides	None	Invariant Risk Minimization (Arjovsky et al., 2019)
December 7	Counterfactuals Mediation	Video Slides	Chapter 14 of ICI (Coming soon)	Identifiability of Path-Specific Effects (Avin, Shpitser, & Pearl, 2005)

Course Mailing List

Sign up for the course mailing list to receive updates about the course:

Course Textbook

Draft of first 10 chapters (continually updated with new chapters throughout the course):

This is a book draft, so I greatly appreciate any feedback you’re willing to send my way. If you’re unsure whether I’ll be receptive to it or not, don’t be. Please send any feedback to me using the “Book” option of the feedback form. Feedback can be at the word level, sentence level, section level, chapter level, etc. Here’s a non-exhaustive list of useful kinds of feedback:

• Typoz.
• Some part is confusing.
• You notice your mind start to wonder or don’t feel motivated to read some part.
• Some part seems like it can be cut.
• You feel strongly that some part absolutely should not be cut.
• Some parts are not connected well.
• When moving from one part to the next, you notice that there isn’t a natural flow.
• A new active reading exercise you thought of.

Prerequisites

There is one main prerequisite: basic probability. This course assumes you’ve taken an introduction to probability course at the undergraduate level or have had equivalent experience. Topics from statistics and machine learning will pop up in the course from time to time, so some familiarity with those will be helpful, but is not necessary. For example, if cross-validation is a new concept to you, you can learn it relatively quickly at the point in the course that it pops up. And in Section 2.4 of the book, we give a primer on some statistics terminology that we’ll use.

FAQs

Q: Where should I ask questions about a given lecture?
A: Use the YouTube comment selection below the relevant video. I check it once per day on week days.

Q: Is this course for credit?
A: No.

Q: Is this course free?
A: Yes!

Q: What time is the course?
A: Only the guest talks will have specific times (listed in the schedule). The regular lecture videos won’t be live and will usually be uploaded to YouTube on Mondays.

Q: I’m not receiving course emails.
A: Email me with “[Causal Course]” at the beginning of your email subject, and I’ll fix it.

Feedback

If you have any feedback about the course to send my way, I welcome it! Please send it here. You can include your name or not include your name. Either works.

Potential Reading Group Papers by Week

We will have a small weekly reading group that runs in parallel to the course. Before any given week’s reading group meeting, 1-3 people will have read the week’s paper in detail and already thought about discussion topics. These 1-3 people will then lead a discussion of a small number of people who have all made themselves familiar with the paper. The discussion group will be kept small (at most 15) in order to facilitate quality discussion. You can ensure that you have a place in the discussion group every week you’d like by signing up to be a discussion leader for at least one week. Below, I give a list of potential reading group papers, organized by week/topic, just like the course schedule is. You can email me at bradyneal11@gmail.com to let me know that you’d like to lead a certain week’s discussion, which paper(s) you’re considering, or to discuss other papers you’d like to discuss that are not on the list.

1. Motivation and Preview - No reading group
2. Potential Outcomes
   • Does obesity shorten life? The importance of well-defined interventions to answer causal questions (Hernán & Taubman, 2008)
   • Does Obesity Shorten Life? Or is it the Soda? On Non-manipulable Causes (Pearl, 2018)
3. Graphical Models and SCMs
   • On the Interpretation of do(x) (Pearl, 2019)
   • Quantifying causal influences (Janzing et al., 2012)
   • Trygve Haavelmo and the Emergence of Causal Calculus (Pearl, 2014)
4. Randomized Experiments, Frontdoor Adjustment, and do-calculus
   • Single World Intervention Graphs: A Primer (Richardson & Robins, 2013)
   • The Paper of How: Estimating Treatment Effects Using the Front-Door Criterion (Bellemare & Bloem, 2019)
   • On Pearl’s Hierarchy and the Foundations of Causal Inference (Bareinboim et al., 2020)
5. Estimation and Conditional Average Treatment Effects
   • Estimating individual treatment effect: generalization bounds and algorithms (Shalit, Johansson, & Sontag, 2017)
   • Adapting Neural Networks for the Estimation of Treatment Effects (Shi, Blei, Veitch, 2019)
   • Generalized Random Forests (Athey, Tibshirani, Wager, 2019)
   • Meta-learners for Estimating Heterogeneous Treatment Effects using Machine Learning (Künzel et al., 2017) (caution: not about meta-learning in the ML sense)
6. Sensitivity Analysis
   • Making sense of sensitivity: extending omitted variable bias (Cinelli & Hazlett, 2019)
   • Sense and Sensitivity Analysis: Simple Post-Hoc Analysis of Bias Due to Unobserved Confounding (Veitch & Zaveri, 2020)
   • An Introduction to Sensitivity Analysis for Unobserved Confounding in Non-Experimental Prevention Research (Liu, Kuramoto, & Stuart, 2013)
   • Sensitivity Analysis of Linear Structural Causal Models (Cinelli et al., 2019)
7. Instrumental Variables, Regression Discontinuity, Difference-in-Differences, and Synthetic Control
   • Improving Causal Inference: Strengths and Limitations of Natural Experiments (Dunning, 2007)
   • Alternative Causal Inference Methods in Population Health Research: Evaluating Tradeoffs and Triangulating Evidence (Mattay et al., 2019)
   • Deep IV: A Flexible Approach for Counterfactual Prediction (Hartford et al., 2017)
   • Regression Discontinuity Designs in Economics (Lee & Lemieux, 2010)
   • Synthetic Controls (there are several different Abadie papers; message me, if you’re interested in this topic)
8. BREAK
9. Causal Discovery without Experiments
   • Inferring causation from time series in Earth system sciences (Runge et al., 2019)
   • Distinguishing Cause from Effect Using Observational Data: Methods and Benchmarks (Mooij et al., 2016)
   • Do-calculus when the True Graph Is Unknown (Hyttinen, Eberhardt, Jarvisalo, 2015)
   • Review of Causal Discovery Methods Based on Graphical Models (Glymour, Zhang, & Spirtes, 2019)
   • Causal inference by using invariant prediction: identification and confidence intervals (Peters, Bühlmann & Meinshausen, 2016)
   • Nonlinear causal discovery with additive noise models (Hoyer et al., 2008)
   • Causal Discovery from Heterogeneous/Nonstationary Data with Independent Changes (Huang et al., 2020)
10. Causal Discovery with Experiments
    • Experiment Selection for Causal Discovery (Hyttinen, Eberhardt, Hoyer, 2013)
    • Characterization and Greedy Learning of Interventional Markov Equivalence Classes of Directed Acyclic Graphs (Hauser & Bühlmann, 2012)
    • Characterizing and Learning Equivalence Classes of Causal DAGs under Interventions (Yang, Katcoff, & Uhler, 2018)
    • Joint Causal Inference from Multiple Contexts (Mooij, Magliacane, & Claassen, 2020)
11. Transportability and Transfer Learning
    • External Validity: From Do-Calculus to Transportability Across Populations (Pearl & Bareinboim, 2014)
    • A causal framework for distribution generalization (Christiansen et al., 2020)
    • Causal inference and the data-fusion problem (Bareinboim & Pearl, 2016)
    • On Causal and Anticausal Learning (Schölkopf et al., 2012)
    • Domain Adaptation under Target and Conditional Shift (Zhang et al., 2013)
    • Multi-Source Domain Adaptation: A Causal View (Zhang, Gong, & Schölkopf., 2015)
    • Invariant Models for Causal Transfer Learning (Rojas-Carulla et al., 2016)
    • Domain Adaptation As a Problem of Inference on Graphical Models (Zhang et al., 2020)
    • Domain Adaptation by Using Causal Inference to Predict Invariant Conditional Distributions (Magliacane et al., 2018)
12. Counterfactuals, Mediation, and Path-Specific Effects
    • Identification, Inference and Sensitivity Analysis for Causal Mediation Effects (Imai, Keele, & Yamamoto, 2010)
    • Identifiability of Path-Specific Effects (Avin, Shpitser, & Pearl, 2005)
    • Interpretation and Identification of Causal Mediation (Pearl, 2014)
13. TBD - Overflow Week
14. Causal Representation Learning
    • Visual Causal Feature Learning (Chalupka, Perona, & Eberhardt, 2015)
    • Discovering causal signals in images (Lopez-Paz et al., 2017)
    • Invariant Risk Minimization (Arjovsky et al., 2019)