Engage live with a leading expert and cohort of experienced professionals
Designed for technology professionals looking to go from 0 to 100 in
Causal Inference
Data Scientists
looking to understand the important distinction between prediction and causal inference
Machine Learning Engineers
who want an introduction to the core material that supports more contemporary machine learning on causal inference



Product Managers
who want guidance on interpreting causal analysis and directions for conducting it



Meet your
Instructor
Prof. Scott Cunningham
Author of Causal Inference:
The Mixtape
Professor at Baylor University


Scott Cunningham is a Professor of Economics at Baylor University. He is also the author of the recent book on causal inference, Causal Inference: The Mixtape. He is in high demand as a teacher of causal inference and has taught several dozen workshops on the subject with large scale enterprises (including Meta, Twitch, Etsy), governments and universities all over the world.
He has published in top journals of economics and is passionate about teaching causal inference to as many as he can because he believes strongly it can help people better understand the impact of our private and public choices on society.
He did his PhD in economics from the University of Georgia in 2007 and has been at Baylor ever since. His research is in health and labor economics like mental health care, corrections, suicide, sex work, violence against women, abortion policy and drug policy.
.png)
Meet your
Instructor
Prof. Scott Cunningham

Prof. Scott Cunningham
Author of Causal Inference: The Mixtape

Professor at Baylor University

Scott Cunningham is a Professor of Economics at Baylor University. He is also the author of the recent book on causal inference, Causal Inference: The Mixtape. He is in high demand as a teacher of causal inference and has taught several dozen workshops on the subject with large scale enterprises (including Meta, Twitch, Etsy), governments and universities all over the world.
He has published in top journals of economics and is passionate about teaching causal inference to as many as he can because he believes strongly it can help people better understand the impact of our private and public choices on society.
He did his PhD in economics from the University of Georgia in 2007 and has been at Baylor ever since. His research is in health and labor economics like mental health care, corrections, suicide, sex work, violence against women, abortion policy and drug policy.

Accompanied by Guest Instructors with experience from
Leading Companies

Brian Quistorff
Ex-Senior Researcher @ Microsoft
.png)
Pedro Sant'Anna
Senior Researcher @ Microsoft
What you get out of
This Course



About Scott’s
Live Cohort
For most of human history, scientists worked with small amounts of data by today’s standards. But that is no longer the case – today we are drowning in data, and everyone is expected to know how to use it. This course is important because it lays out the difference between prediction and causal inference, and focuses on the latter. This will help decision makers better understand when the tool calls for a causal question to be answered and when it calls for a predictive answer. Failure to make such distinctions can lead to tragic mistakes in judgment.
The modern field of causal inference is a blend of statistics and econometrics built on a theory of counterfactuals and the potential outcomes framework. This course is primarily based on the design tradition whose architects were awarded the 2021 Nobel Prize in economics (David Card, Josh Angrist and Guido Imbens), as opposed to the model-based tradition of causal inference found in Judea Pearl’s work, though they are complements not substitutes.
This course trains you in the art and science of causal inference by studying closely the main research designs within this field – regression discontinuity, instrumental variables, difference-in-differences and synthetic control – as well as the estimators that are often best suited for them. It will also advance your knowledge through spending time going over applications in R and python.
The goals and learning objectives of this course are to develop a solid understanding of:

Design based causal inference, including the major research designs of regression discontinuity, difference-in-differences, instrumental variables and synthetic control

Practically implementing these designs in R and python, including interpreting output

Session 1 - Potential outcomes
We will learn the potential outcomes model which provides the mathematical and statistical notation for all design-based causal inference. Specifically, we will:
.png)
Use this notation to lay out key aggregate causal parameters like the average treatment effect,
.png)
Illustrate the source of selection bias using a decomposition of simple comparisons and note the problems created when people choose treatments based on expected personal gains
.png)
Show why randomization is such an important feature in causal inference
Sessions 2 & 3 - Diff-in-Diff
The most popular quasi-experimental method in design-based causal inference is the difference in differences. The intuition and setup are simple: a group of treated units are observed after treatment and before treatment which are then compared to another group of untreated units after and before. In these three sessions, we will:
.png)
Learn models developed to correctly obtain unbiased estimates of the average treatment effect on the treated units under a variety of common situations (e.g., differential timing, controlling for covariates)
.png)
Develop a deeper understanding of how one might obtain some indirect evidence that the method’s key assumptions are reasonable to make
.png)
Implement these methods using code and data


.png)
Sessions 4 & 5 - Synthetic Control
Susan Athey, former Chief Economist atMicrosoft and John Bates Clark award winner, and Guido Imbens, the 2001 co-recipient of the Nobel Prize, have called synthetic control one of the most important innovations in causal inference of the last twenty years. In this session, we will:
.png)
Learn the traditional method developed by Alberto Abadie and coauthors,
.png)
Practice its implementation using code and sample data
.png)
Provide some guidance about ways to correctly augment the method when certain challenges present themselves

Sessions 6 - Regression Discontinuity
Inferring causality always involves understanding how people or things are moved into and out of some intervention. In the A/B test, that involves randomization. With regression discontinuity designs, it involves a score. When people or things possess a score over some cutoff, they are moved into an intervention and regression discontinuity formalizes this process so that we can obtain valid estimates of the causal effect of the intervention on outcomes we care about. In this session, we will:
.png)
Learn the estimation techniques that are essential to obtaining causal effects when treatment assignment is based on scores as opposed to randomization
.png)
Practice RDD using code and large administrative data from a government agency
.png)
Provide suggestions about data visualization

Don’t miss this unique opportunity to
Accelerate Innovation In
Your Organization
Based on workshops conducted with





