Observation and Experiment: An Introduction to Causal Inference

by Paul R. Rosenbaum

In the daily news and the scientific literature, we are faced with conflicting claims about the effects caused by some treatments, behaviors, and policies. A daily glass of wine prolongs life, or so we are told. Yet we are also told that alcohol can cause life-threatening cancer and that pregnant women should abstain from drinking. Some say that raising the minimum wage decreases inequality while others say it increases unemployment. Investigators once confidently claimed that hormone replacement therapy reduces the risk of heart disease but today investigators confidently claim it raises that risk. How should we study such questions?

Observation and Experiment is an introduction to causal inference from one of the field’s leading scholars. Using minimal mathematics and statistics, Paul Rosenbaum explains key concepts and methods through scientific examples that make complex ideas concrete and abstract principles accessible.

Some causal questions can be studied in randomized trials in which coin flips assign individuals to treatments. But because randomized trials are not always practical or ethical, many causal questions are investigated in nonrandomized observational studies. To illustrate, Rosenbaum draws examples from clinical medicine, economics, public health, epidemiology, clinical psychology, and psychiatry. Readers gain an understanding of the design and interpretation of randomized trials, the ways they differ from observational studies, and the techniques used to remove, investigate, and appraise bias in observational studies. Observation and Experiment is a valuable resource for anyone with a serious interest in the empirical study of human health, behavior, and well-being.

Genres: Mathematics, Science, Nonfiction

374 pages, Hardcover

Published August 14, 2017

About the author

Paul R. Rosenbaum is the Robert G. Putzel Professor Emeritus of Statistics and Data Science at the Wharton School of the University of Pennsylvania. He is the author of Observation and Experiment: An Introduction to Causal Inference, Design of Observational Studies, Observational Studies, and Replication and Evidence Factors in Observational Studies.

Reviews

[Kyle · Rating 5 out of 5]

This is an excellent introduction and explanation of how to use statistics to reason. I learned a lot about how one should design experiments so that you can know how strong of a conclusion you can make from the data.

As a person who doesn't mind the math, I was a little disappointed that the math isn't included, but Rosenbaum always has a reference to look up for more details, so that my disappointment was mollified. The advantage of this is getting to read Rosenbaum's lucid explanations in "plain English". He does an excellent job of explaining things with good examples, and making sure that the limits of the examples are established.

I was somewhat familiar with many of the concepts, but Rosenbaum does a great job of explaining the though processes for randomized experiments, natural experiments, observational studies, how to produce matches for treatment vs control, and (most interesting for me) how to look at the sensitivity of these studies to bias. Design sensitivity is extremely well explained and something extremely important. Essentially knowing the robustness of the evidence is always difficult and sensitivity analysis lets you get at this.

I would recommend this to anyone who has an interest in statistical methods. It is very accessible and Rosenbaum is a very good writer. Whether or not you like math, the book is very good at explaining its concepts and elucidating them with examples. Its examples are heavily tilted toward the medical field, but they are not exclusive to it by any means. Overall, just a very good introduction and explanation of concepts.

[Michiel · Rating 3 out of 5]

I have some mixed feelings about this book. The first part about randomization was superb, truly building the theory from the ground up. The latter chapters were sometimes hard to follow because they are inevitably more complex. The author chooses to keep the number of equations at a minimum, preferring examples and tables. In my opinion, some simulation examples might be more didactic. The summary and lists of terms at the end, however, make this book well worth adding to one's library.

[John Mihelic · Rating 4 out of 5]

I have to point out that there are people who are critical of Randomized Control Experiments (RCEs) Like Professors Ziliak and McCloskey, critics who get a passing mention here in the beginning but not by name. So when the RCEs are held up as the gold standard of understanding causation in this book, I'm a little circumspect. Overall though, it is a very accessible book about how we can pull apart causation using the tools most widely used in the academy.

[Tripp · Rating 4 out of 5]

It's a well-written book that clearly explains why and when we can draw valid causal inference. He has a couple good chapters devoted to observational data and how you should think about control groups and counterparts.

[Lee Richardson · Rating 4 out of 5]

This book starts off great. The chapter on why randomization works, and the chapter that explains the logic of propensity scores, are awesome. For these two chapters, I'd recommend this book.

But I'd also recommend skimming the later chapters (at most). It felt like an academic presentation, where the speaker goes on and on, and no one in the room is following what they're saying anymore.

And is kind of a shame. It's such an important topic, and a concise, better written book could have had a major impact.

[Dylan Armbruster · Rating 5 out of 5]

Discover the inner workings of randomization, delve into the intricacies of propensity scores, and explore the realm of observational studies through the guidance of a true expert, Rosenbaum, who not only possesses profound knowledge but also excels in the art of effective teaching. Within the pages of this book, Rosenbaum, drawing from his own experiences as a student, ensures that your journey is accompanied by unwavering support. Your hand will be held throughout, a gesture that is genuinely appreciated. However, if you find yourself well-versed and confident, rest assured that you can navigate this text independently. To cater to various levels of expertise, each chapter concludes with more technical sections. The book stands as a self-contained reservoir of knowledge. Furthermore, Rosenbaum has structured this book, like his others, in an intriguing manner. Readers have the flexibility to embark on their exploration from various entry points, a testament to his unique approach.

[Suzanne · Rating 3 out of 5]

3.5 stars. An interesting read for someone who does mostly observational, correlations, and longitudinal research. But the book focuses on distinct group comparison (treated v. not treated, experienced event v. didn’t). I wish it had gone one step farther to observed differences that are not dichotomous.

[Yi · Rating 5 out of 5]

An excellent introduction with little technical details.

[Monika]

CWRU

[Martine · Rating 4 out of 5]

Very interesting read, some examples could be presented more didactically, it is oke to sometimes use a formula. You need two bookmarks to read it, one for the text and one for the excellent notes.