Causal Inference for Data Science

When you know the cause of an event, you can affect its outcome. This accessible introduction to causal inference shows you how to determine causality and estimate effects using statistics and machine learning.

A/B tests or randomized controlled trials are expensive and often unfeasible in a business environment. Causal Inference for Data Science reveals the techniques and methodologies you can use to identify causes from data, even when no experiment or test has been performed.

In Causal Inference for Data Science you will learn how to:

• Model reality using causal graphs
• Estimate causal effects using statistical and machine learning techniques
• Determine when to use A/B tests, causal inference, and machine learning
• Explain and assess objectives, assumptions, risks, and limitations
• Determine if you have enough variables for your analysis

It’s possible to predict events without knowing what causes them. Understanding causality allows you both to make data-driven predictions and also intervene to affect the outcomes. Causal Inference for Data Science shows you how to build data science tools that can identify the root cause of trends and events. You’ll learn how to interpret historical data, understand customer behaviors, and empower management to apply optimal decisions.

About the technology

Why did you get a particular result? What would have lead to a different outcome? These are the essential questions of causal inference. This powerful methodology improves your decisions by connecting cause and effect—even when you can’t run experiments, A/B tests, or expensive controlled trials.

About the book

Causal Inference for Data Science introduces techniques to apply causal reasoning to ordinary business scenarios. And with this clearly-written, practical guide, you won’t need advanced statistics or high-level math to put causal inference into practice! By applying a simple approach based on Directed Acyclic Graphs (DAGs), you’ll learn to assess advertising performance, pick productive health treatments, deliver effective product pricing, and more.

What's inside

• When to use A/B tests, causal inference, and ML
• Assess objectives, assumptions, risks, and limitations
• Apply causal inference to real business data

About the reader

For data scientists, ML engineers, and statisticians.

About the author

Aleix Ruiz de Villa Robert is a data scientist with a PhD in mathematical analysis from the Universitat Autònoma de Barcelona.

Table of Contents

Part 1
1 Introducing causality
2 First steps: Working with confounders
3 Applying causal inference
4 How machine learning and causal inference can help each other
Part 2
5 Finding comparable cases with propensity scores
6 Direct and indirect effects with linear models
7 Dealing with complex graphs
8 Advanced tools with the DoubleML library
Part 3
9 Instrumental variables
10 Potential outcomes framework
11 The effect of a time-related event
A The math behind the adjustment formula
B Solutions to exercises in chapter 2
C Technical lemma for the propensity scores
D Proof for doubly robust estimator
E Technical lemma for the alternative instrumental variable estimator
F Proof of the instrumental variable formula for imperfect compliance

Aleix Ruiz de Villa is a freelance data science consultant with a PhD in mathematical analysis from the Universitat Autonoma de Barcelona. Aleix has worked in the journalism, retail, transportation and software development industries. He is the founder of the Barcelona Data Science & Machine Learning Meetup.