The Book of Why

The Book of Why

The New Science of Cause and Effect

Book - 2018
Average Rating:
Rate this:
2
"Everyone has heard the claim, "Correlation does not imply causation." What might sound like a reasonable dictum metastasized in the twentieth century into one of science's biggest obstacles, as a legion of researchers became unwilling to make the claim that one thing could cause another. Even two decades ago, asking a statistician a question like "Was it the aspirin that stopped my headache?" would have been like asking if he believed in voodoo, or at best a topic for conversation at a cocktail party rather than a legitimate target of scientific inquiry. Scientists were allowed to posit only that the probability that one thing was associated with another. This all changed with Judea Pearl, whose work on causality was not just a victory for common sense, but a revolution in the study of the world"-- Provided by publisher.
Publisher: New York, NY : Basic Books, [2018]
Characteristics: x, 418 pages : illustrations ; 25 cm
Edition: First edition
Copyright Date: ©2018
Multiscript Copyrightdate: 018
ISBN: 9780465097609 (hardcover)
046509760X (hardcover)
Call Number: 501 PEARL

Opinion

From the critics


Community Activity

Comment

Add a Comment
t
tj_is_cool
Nov 18, 2018

Very long and very boring presentation.

g
gaetanlion
Sep 07, 2018

Really unclear. I read his book on “Causality” and did not understand much of it despite a fair math background. Now, Pearl writes a book for a general audience. But, it is still a hard read. And, I found it disappointing on several counts.

Pearl’s tone is not balanced. He presents Sewall Wright, the founder of path analysis, as an ignored martyr. However, that did not seem to be the case. So, what’s the beef? Pearl’s tone often resembles the one of either Nassim Nicholas Taleb or Benoit Mandelbrot (a bit too belligerent than called for).

Pearl way overstates his case. He indicates that without his methods (diagrams, and do-calculus) you can’t figure out if you double the value of X what will happen to Y? Or if you have three different independent variables (Xs), what are their respective influence or impact on Y? These relate to his 2nd and 3d level of causation. But, we do that with regressions, Granger Causality, VAR, factor analysis, etc. I am not saying these methods would address all his questions on the 2nd and 3d level. But, they answer many. And, the remaining ones could be addressed through path analysis (that is not as ignored as he suggests).

His explanations are cryptic. His chapter 6 on math paradoxes describe perplexing math conundrums. But, his explanations fail to clarify.

The book has quite a few mistakes. On page 51 he mentions vector autocorrelation instead of vector autoregression. On page 62, he confuses correlation coefficient with regression coefficient. His limitations of regression (pg. 286) are way too restrictive. He also states that “linear models do not allow interaction” (pg. 322). However, regression can capture non-linear effects and interactions just fine.

When he compares models, his explanations are unclear. He contrasts a regression model with a structural equation model (SEM) and a structural causal model (SCM) (pg 274 – 279). And, he considers them bad, better, and best respectively. But, the explanation of the differences between them is differentiated by trivial semantics.

The book is not all bad. He makes some good points. The first one is that the old adage “correlation does not entail causation” is wrongly treated as a universal rule. Pearl is points out that the fathers of statistics (Galton, Spearman, Pearson, Fisher, Gauss) have way overstated this case (correlation does not entail…). Many relationships between variables (correlation) have an explicit direction (causation). Given that, observational studies could be of huge value to medical research as they are cheaper and easier to implement than randomized controlled trials. Also, the last 20 pages of the book are the best. He mentions the work of Elias Bareinboim that facilitates determining the relevance of prior studies to study a new situation and also to assess the implication of selection bias within such studies. This is incredibly cool stuff. He later fleshes out the implication and prospect of Artificial Intelligence endowed with an understanding of causality and a capability for moral judgment (distinguishing right from wrong). That is also very interesting material.

However, in conclusion, I can’t recommend this book for the mentioned reasons. Instead, I suggest you study the basics of path analysis. You will get more practical understanding in less time. That’s been my own experience.

Age

Add Age Suitability

There are no ages for this title yet.

Summary

Add a Summary

There are no summaries for this title yet.

Notices

Add Notices

There are no notices for this title yet.

Quotes

Add a Quote

There are no quotes for this title yet.

Explore Further

Browse by Call Number

Recommendations

Subject Headings

  Loading...

Find it at JCPL

  Loading...
[]
[]
To Top