Despite a reputation for being worse than damned lies, statistics conquered its enemies during the 20th century and now reigns over science. What happens in a scientific experiment is considered unworthy of serious consideration (that is, you can’t get it published) unless formulas determine that the result is “statistically significant.”
Those formulas involve all sorts of obscure concepts such as the central limit theorem, correlation coefficients, standard deviation and regression analyses — all involving mathematical wizardry that would be relegated to books in the restricted section of the Hogwarts library. But Wheelan, a Dartmouth economist, has provided a volume offering safe reading for all ages. With humor and an engaging conversational style, he walks the reader through the basics of statistical concepts and their applications, using real-world examples to illustrate how statistics work and why they matter.
All in all, it’s an excellent book. But with subtle weaknesses. Wheelan ignores Bayesian statistics, which is often the more appropriate approach for dealing with many of the issues he discusses. And in a chapter toward the end (called a “warning label” about common statistical mistakes), he discloses that the methods he has been presenting often don’t work. “A shocking amount of expert research turns out to be wrong,” he writes. Yet throughout most of the book he presents standard statistics as the way to get right answers. In many cases, however, statistics stripped of its
mysteries turns out to be an emperor with no clothes.
W.W. Norton & Co., 2013, 282 p., $26.95