Week 10 Info: Causal Inference
So far in this course, we have mostly been focused on predictive modeling, though when we discussed linear models we talked about how in certain circumstances the coefficients of a linear regression model could be interpreted as causal effects. This week, we are going to talk about a rapidly growing field called Causal ML, which will be a window for us to apply machine learning to Causal Inference problems. Causal Inference is important when you are trying to use models to determine what will happen if you impact the data generating process in some way, such as offering a sale or other incentives to customers or giving a treatment to a patient suffering from a condition. These questions are of great practical importance, but are challenging to answer outside the context of controlled-randomized experiments. Indeed A/B testing is an extremely important domain for a large number of companies, but it is not always possible or ethical to run experiments. For these situations, a host of methods have been developed in recent years to apply machine learning tools to learn causal effects. This week we will learn about causal inference, the basic causal estimands that you learn when fitting a causal model, and how machine learning has led to the newest causal inference techniques. The methods used for CausalML are generally tree-based, including BART (which it turns out we are not covering) and Causal Forests, which we will learn about this week.
Just a note, Spring Break begins in the middle of this week and so the typical class schedule will pause. There will be no meetup next week and we will return the week after.
You can read more about the plan for the week and the reading/resources in Module 10.
Lab 6 due Sunday, April 19th at midnight.