Causal Inference in Python

Chapter 3. Graphical Causal Models

In Chapter 1 you saw how causal inference can be broken down into two problems: identification and estimation. In this chapter, you’ll dive deeper into the identification part, which is arguably the most challenging one. This chapter is mostly theoretical, as you will be playing with graphical models without necessarily estimating their parameters with data. Don’t let this fool you. Identification is the heart of causal inference, so learning its theory is fundamental for tackling causal problems in real life. In this chapter, you will:

• Get an introduction to graphical models, where you will learn what a graphical model for causality is, how associations flow in a graph, and how to query a graph using off-the-shelf software.

• Revisit the concept of identification through the lens of graphical models.

• Learn about two very common sources of bias that hinder identification, their causal graph structure, and what you can do about them.

Thinking About Causality

Have you ever noticed how those cooks in YouTube videos are excellent at describing food? “Reduce the sauce until it reaches a velvety consistency.” If you are just learning to cook, you have no idea what this even means. Just give me the time I should leave this thing on the stove, will you! With causality, it’s the same thing. Suppose you walk into a bar and hear folks discussing causality (probably a bar next to an economics department). In that case, you will hear them say how the confounding ...