Stephen Campbell's Flaws and Fallacies in Statistical Thinking was originally published in 1974. It is now available in a variety of formats, ranging from a Dover reprint to Kindle, iTunes, and OpenLibrary. The author provides a very readable account of many kinds of flawed statistical reasoning (more than 20 by a casual count). He illustrates each with examples that are, for the most part, drawn from real life. In some cases the fallacies and flaws are inadvertent, while others are likely quite intentional. The text reads easily, with only a few typographical errors noted.
The book is intended to help the non-statistician recognize abuses of statistical reasoning that appear in public discourse; it accomplishes this. Today, many decision-makers are asked to provide data to support their conclusions; this book will help them think carefully about the conclusions they choose to draw and the statistical evidence they choose to provide to support their conclusions. There are very few equations in the book, and because those that appear are relegated to footnotes, the book is quite accessible. The topics covered are listed in the table of contents that accompanies this review. The cartoons sprinkled throughout the book are amusing, though dated.
A new edition of the book would be welcome. In 1974, this would have been a useful supplemental text in a general education mathematics course. Few classes today, however, would appreciate the adoption of a supplemental text that is forty years old, with examples dating from decades before the original publication date of the book. Instead, students would likely be distracted or offended by evidence of the social changes that have occurred in the past forty years, especially with respect to race, ethnicity, and gender. (One particularly troublesome example is a Playboy interview from 1966, in which the commander of the American Nazi Party uses racial slurs when discussing his views on I.Q. scores of different races.) While I often teach a general education course that would benefit from the statistical topics discussed, I would not find it acceptable today to use this as a supplemental text. Directed readings from it would also have to be selected with care.
Joel Haack is Professor of Mathematics at the University of Northern Iowa.