Statistics, admittedly it's not everyone's favourite subject. In fact I have a friend who insists on referring to topic as sadistics so it's probably a least favourite subject for some. However a staggering amount of life revolves around statistics from politicians throwing numbers at each other in the House of Commons to analysing cost / benefit ratios for new drugs. In fact I don't think it's hyperbole to say that without a decent grounding in statistics you are at a disadvantage in day to day life. Despite working in statistics-heavy research for several years my knowledge is somewhat erratic as I have picked it up more or less as needed - so Statistics In A Nutshell to the rescue!
First impressions of the book were, well, mixed. I'm used to Nutshell books being incredibly terse and this one has blocks of text - even well written descriptions of how to use the different statistical methods. What you have here is really a hybrid between the "revision notes" approach typically taken by Nutshell and a full on text book. While it would be possible to use it as a textbook I wouldn't recommend it simply because the sheer information density would have you running for the aspirin before too many pages. The additional prose is an absolute godsend when using the book as a reference as it describes not only statistical methods but their uses too.
Content wise Statistics In A Nutshell covers pretty much all the bases you might hope for. The first five chapters cover the fundamentals of measurement, probability, data management, descriptive statistics and research design. If you are only interested in presenting data summaries, demonstrating trends and the like then you probably don't actually need to venture much further into the book but you would be missing out on a lot if you ignored it! In particular Chapter six on critiquing statistics presented by others is an absolute gem. Statistics do seem to be more frequently used to confuse and obfuscate rather than inform and enlighten and do have the useful, for the devious statistician at least, quirk that by some careful tailoring they can appear to show results that are actually entirely unfounded. This chapter takes you for a walk through some of the most frequently seen flaws in statistical representations to enable you to gain a more accurate understating of the truth behind what is being represented, or misrepresented.
The remainder of the book is aimed at more hard-core statistician, covering various flavours of inferential statistics, parametric vs. non parametric approaches, linear regression, correlation and variance analysis. Each topic is covering a clear and concise way with examples showing where and how the techniques would be used. The final three chapters of the book are given over to statistics in specialist areas - business, medical and educational statistics. These chapters introduce or expand on techniques which are particular to the field. There are also a number of appendices which recap basic mathematics (very useful if like me it's a long time since you were last in a maths class), introduce common statistical packages and finally provide an excellent reference section.
All in all I would say that Statistics In A Nutshell is an excellent book that actually delivers more than it promises and would be a worthy desktop addition to anyone who works with statistics on an at least semi-regular basis.
I came to the book looking for a refresher on university level statistics (after a 20 year absence). It is different from the other stats books that I had in that the explanations are more geared towards use rather than proving everything from first-principles. As such it met my needs very well.
It's not a basic guide and if you're scared of stats I doubt it is the right choice. If, however, you need more confidence on which approaches to use given the type of problem you face it is excellent.
For example, coverage of the Kruskal Wallis H test. * Gives the wrong formula: n instead of N, and with (mean R-(n+1))/2 squared rather than (mean R-(N+1)/2)^2. * Gives Chi Square df for H tests as 8, rather than k-1 (the example has k=3). * Tells the reader that an H of 2.26 is enough to reject the null hypothesis with a critical Chi Square value of 15.51.
Previous to this there are a few typos, but this is just plain wrong, and is liable to seriously derail anyone attempting to use this book to learn about statistics. In comparison, while it covers less ground, I'd have to say Perry Hinton's Statistics Explained is a better book. Covers all the basic tests, gives better examples and goes into the maths a bit more while still managing to be more accessible.