Joel Best

Short and not very deep book about when to be skeptical of statistics. I think the message could have been given even more briefly: often, statistics are unreliable because of problems of data collection and definition of terms, whether in surveys or in other ways. Because of innumeracy and natural tendencies to accept claims that correspond to our own preexisting inclinations, we too rarely question statistics—even ones that are unbelievable if you think about them for a second, like show more “over a hundred thousand women die from anorexia in the US each year,” which conflates suffering and dying. Best advocates that we ask ourselves how a particular statistic was created and with what reasons when judging its reliability, because neither total cynicism nor total credulity is justified. Maybe a book I’d give a kid in high school. show less

A worthy sequel to "Damned Lies and Statistics" in much the same vein.

The book has some significant flaws, however. The statement that there is no pro-bullying lobby is actually false; some of the more conservative politicians these days have stated pretty clearly that bullying is the kind of treatment queer kids deserve. There is a really awkward and unhelpful bit about percentages that should have been improved or omitted.

The book's strengths are the excellent writing and organization, show more particularly the way different kinds of statistics are broken down into missing numbers (relevant numbers that are simply overlooked in the discussion, sometimes because nobody likes them, sometimes because they are historical and forgotten), confusing numbers (more or less reliable numbers that are presented in misleading ways through arithmetical or graphical tricks or errors), scary numbers (numbers which are meant to alarm and motivate or intimidate), authoritative numbers (numbers from unusually reliable sources that may yet be misinterpreted), magical numbers (numbers on which important decisions rest that are amenable to manipulation) and contentious numbers (numbers that pop up in the course of a stats war, usually arising where there is little strong consensus on mechanisms, so that all arguments rest on statistical correlation or lack of it).

There is an afterword w/ more resources. The resources seem to be of variable quality; Numberwatch appears to be a mere hobby-horse for some pompous old professor, but StatLit looks kind of useful. show less

Statistics are constructed not discovered. It is people that create statistics, consequently statistics are coloured by the worldview of the constructor. There is no such thing as a neutral statistic. This book provides ample evidence that statistics are the products of social activity. Joel Best, a professor of sociology and criminal justice, has produced a readable and informative book. He draws upon statistics from many walks of life to illustrate his points.

Those who create statistics show more want to prove something. Best, shows how by asking key questions we can help expose these underlying perspectives:

Who created the statistic?
Why was this statistic created?
How was the statistic created?

Best deals with key issues such as the source of bad statistics: bad guesses, descriptive definitions, confusing questions and biased samples; mutant statistics: the ways in which good statistics can be mangled, misused and misunderstood; and the problems of statistical comparison. It requires no background in maths or statistics. It will help readers be more critical in their acceptance of statistical data. show less

This is a book about the social context surrounding the creation and dissemination of social statistics. It is in no way a book about the mathematical discipline of statistics, which is generally applied in textbooks to widgets, not people. Interestingly, the word statistics is derived from the word state, "statists" were people who collected and analyzed data about the state. Eventually, we get statisticians who may collect and analyze data about just about anything.

The book discusses the show more motivations behind the creation of statistics in a fairly un-judgmental way, avoiding laying too much stress on the cynical presentation of statistics for the purpose of propaganda.

The book discusses some approaches to statistics that one can take; the first question to ask is "Is that even possible?" It discusses the many things that influence the collection of data. "Organizational practices" are a big factor; if more and more doctors diagnose a child as autistic does that mean that autism is on the rise or just that diagnoses are on the rise?

The statement "suicide is the second leading cause of death among adolescents" is easily satisfied by making the leading cause of death be "not-suicide", so long as at least one adolescent does commit suicide. Not such a useful statistic that one.

Tactics in statistics construction are broadening or narrowing the definition, talking about the "dark figure" (the unreported instances), asking a series of possibly leading questions and yet summarizing all the answers into a single "yes" or "no" response and biased sampling.
Errors in statistics dissemination are transformation of the meaning of a statistic and doing misguided computations, often using possibly wildly incorrect values as inputs. The author calls these sorts of errors mutant statistics.
Comparisons of different values may very well be meaningless when they end up comparing apples and oranges by using different measures, using the same measure at different times, using different definitions, failing to account for differing organizational practices, comparing different populations, and so forth.

In his last chapter but one the author describes several instances of arguments around and about statistics and the construction of different statistics by the opposing parties in the argument to support their side.

A well written, thoughtful book. My new best answer to the quotation of a statistic by anyone will simply be that I need to discover the origin and methodology of the statistic before I can make any comment; unless, of course, the statistic is utterly crazy, like the one with which the author begins the book. Then, I can just cut to the chase. show less