Naked Statistics: Stripping the Dread from the Data - Charles Wheelan
As someone who uses a lot of advanced statistical methods in my profession, I enjoy reading books like this, as they often help me to explain what I am doing to non-scientists. The author is also not a statistician, and that is a good thing, as he approaches the topic from a very practical view-point. His explanations are clear and concise, and technically accurate. It is the best book, by far, for the curious layperson seeking to understand statistics and its use in life.
A few topics were omitted,which puzzle me. Analysis of variance was not covered, which is a shame, especially since it could dove-tail nicely into the discussion around multi-linear regression. More serious was the complete lack of any discussion on Bayesian statistics. As an unabashed Bayesian , I find this to be most unfortunate. There is a huge resurgence in Bayesian methods now, and all serious stats research is Bayesian-based. More importantly, to discuss Bayesian, one must discuss frequentist statistics, including the limitations of the p-value (contrary to what the author would like you to believe, the p-value is not without some serious faults). Here's hoping the author revises the book and includes these important topics, which would improve an already terrific book.