THE ART OF STATISTICS

An exploration of “why we need statistics” and how to use them effectively.

The fact that Darrell Huff’s delightful How to Lie With Statistics (1954) remains in print should convince readers that politicians, demagogues, and advertisers have never had trouble misleading us with numbers and graphs. Still, the study of statistics is widely considered boring, so popular books on the subject work hard to be entertaining; this expert primer mostly measures up. Distinguished British statistician Spiegelhalter (Statistics/Univ. of Cambridge; Sex by Numbers, 2015, etc.), a former president of the Royal Statistical Society, writes that “numbers do not speak for themselves; the context, language, and graphic design all contribute to the way communication is received. We have to acknowledge that we are telling a story.” Some statistics are meaningless—e.g., based on average, a human has one testicle. Some are unhelpful: Vegetarians earn more than meat-eaters, but avoiding meat is unlikely to boost your income. An identical statistic can tell a horror story—e.g., a drug increases the risk of lung cancer by 14%, or not, if it increases the risk from 1 to 1.14 in 1,000. Unlike Huff’s slim volume, Spiegelhalter goes beyond debunking numerical nonsense to deliver a largely mathematics-free but often formidable education on the vocabulary and techniques of statistical science. Almost everyone will understand how “median” differs from “average,” and most will grasp the meaning of a bell curve or that “deduction” (using the rules of logic to come to a conclusion, Sherlock Holmes) is the converse of “induction” (using particular events to draw a general conclusion). Despite careful explanations and a plethora of tables and graphs, readers may strain to understand concepts such as the Poisson distribution, confidence intervals, bootstrapping, or standard deviation, but their efforts will be rewarded.

An admirable corrective to fake news and sloppy thinking.