Keywords: Statistics, data analysis, public policy, current affairs

Title: Breaking The Law Of Averages

Author: William M. Briggs

Publisher: Lulu.com

ISBN: 0557019907


It will be a surprise for most people to discover that the world of statistics, like every other sphere of human life, has deep-seated ideological differences. We all know that economists, climatologists and physicists are prone to falling out with each other, but statisticians? But it is true, at the heart of statistics there are key philosophical disputes about the very meaning of probability. While everyone agrees that the chance of tossing a heads for a fair coin is 0.5, what that actually means will depend on whether you are a frequentist or a Bayesian, and if you're a Bayesian do you tend to the objectivist/logical or subjective schools? Understanding what these mean, and what the implications are as regards statistical thinking, is a key part of this entertaining and thought-provoking little book.

William Briggs is a research scientist, lecturer in statistics and the author of an engaging blog that is always worth seeking out. With this book Briggs sets out to provide the reader with a statistics course that aims to instil statistical thinking rather than providing a set of recipes to be followed. Where other statistics books take great delight in inflicting mind-numbingly tedious step-by-step algorithms on the reader, this book takes advantage of open-source stats software and instead focuses on what the stuff actually means.

This isn't to say that the book avoids math. Far from it, the reader is encouraged to think deeply about what he or she is doing. Statistical thinking is hard and the chances are that following pre-defined formulae is a recipe for getting things wrong. And, time and again, Briggs comes back to these differences between frequentist statistics (which is mostly what gets taught to students), and Bayesian analysis. The follow the steps books actually make fewer demands on the reader, here the reader has to concentrate much more. But that's OK, because Briggs is very good at pointing out where you really need to pay attention (so no slouching at the back there).

The writing is entertaining and keeps the reader engaged, even when the going gets tough. For those who really are following a stats course the book provides exercises and there's a web site to support it. For those who are reading for the hell of it, then there's a lot to be gleaned even if you have no intention of ever carrying out anything more complicated than calculating an average. Along the way Briggs makes plenty of observations about how statistics is done, about the meaning of statistical significance, the publication process in science, how statistics can be used to hide or divert attention and more.

If there's a criticism of the book it's that it really could have done with a better proof-read – sometimes the typos really do get in the way. Hopefully future editions will clear this up. However, that shouldn't distract from what is a very thought-provoking and readable book.

Contents © London Book Review 2009. Published 16 April 2009