Yesterday Judea Pearl, my co-author, sent me a link to a fascinating article that I consider to be even better than a book review. It is something that every author dreams of — positive evidence that our book is making a difference in the real world.
In the article, a group of 47 editors of research journals in the fields of respiratory, sleep, and critical care have compiled “best practices” recommendations for researchers who want to make causal inferences from observational data. That’s a mouthful, so let me explain. Our book, The Book of Why, is all about when you can and cannot answer causal queries on the basis of observational data (*). In short, we argue that if you can formulate an explicit causal model, represented in a simple dot-and-arrow diagram, then there are simple mathematical tests to determine what causal questions you can answer. Questions like “Does this drug affect the likelihood of heart attack?” or “Will working a midnight shift increase my risk of sleep disorders?” In other words, the kinds of questions that patients ask doctors all the time.
The official policy of many journals, including the Journal of the American Medical Association (stated in 2017 in an editor’s blog post) is that authors of medical articles are not allowed to answer such questions. “If it isn’t [a randomized trial], and is a report of an observational study… then all cause-and-effect language must be replaced.” (Emphasis added.)
So you can see the strength of the taboo that Judea and I are challenging. That is why the joint statement of 47 editors from a variety of medical journals is so important. It is a public admission that the guidance of the American Medical Association is outmoded and outdated. We do have the right to use cause-and-effect language, provided that we state our assumptions publicly and transparently.
It would be wrong, of course, for Judea and me to claim credit for this change in attitudes. It has been evolving for some years. However, I think it is very intriguing that Table 2 of the article lists The Book of Why at the top of the suggested resources for people wanting to learn more about causal inference. In fact, there are three books recommended: The Book of Why, Judea’s much more technical book Causality, and a forthcoming (2019) book by Jamie Robins and Miguel Hernan called Causal Inference.
Further, the article says,
We urge authors to consider using causal models when testing causal associations. The scientific, mathematical, and theoretical underpinnings of causal inference, developed by Judea Pearl, James Robins, Miguel Hernan, and others, have evolved sufficiently to permit the everyday use of causal models.
Allowing for the usual opacity of scientific writing, this is just about as ringing an endorsement of Judea’s methods as you could possibly ask for. I encourage anyone who is interested in doing causal analysis of observational studies to take a look at the full article, read our book, and connect with Judea on Twitter.
(*) Here “observational data” is intended to contrast with data gathered from a randomized controlled trial, which has long been the gold standard for establishing causal relationships. (In fact, knee-jerk acceptance of randomized controlled trials is almost as bad as knee-jerk rejection of observational studies. Both of them rely on assumptions that must be scrutinized. But that is a discussion for another time.)