Free online textbooks of statistics.
It is truly amazing how many excellent statistical text "books" are available on the web.
A New View of Statistics. by Will G Hopkins. This is an impressive work that clearly explains many statistical concepts, basic and advanced. It explains things for biologists, with almost no math. The examples are from sports physiology, but any biologist will appreciate the clear illustrated explanations. It is really a superb book, and is frequently updated. My advice is to check out this online book first, and then refer to the others if you can't find what you need.
Hyperstat Online Textbook by David M. Lane (Rice University). A very readable online statistics textbook, with good discussions of ANOVA and probability. Written by a social scientist, so no biological insights. What makes this online text unique is that each chapter also contains an extensive list of links to other resources (articles, online calculators, books, etc.). So even if you don't prefer Lane's style, you will find links to other sources you'll find useful.
Introductory Statistics: Concepts, Models and Applications By David W. Stockburger Southwest Missouri State University. Comprehensive, conventional, well-written statistics text with a behavioral science slant. While it gives a very good explanation of the basics, it doesn't cover ANOVA beyond one-way)or regression beyond simple linear.
Statistics at Square One This very short book by T D V Swinscow provides a concise introduction to statistics. It emphasizes how to do the calculations manually, so explains the math of the tests in detail. It doesn't do quite as good a job of explaining the concepts. But very short, organized, and readable.
Resampling. The New Statistics. This book explains how probability and statistical questions can all be answered by Monte Carlo (resampling) simulations. This approach takes the mumbo-jumbo out of statistics and enables even beginning students to understand completely everything that is done. There is no question that simulations and resampling are very useful for some kinds of analyses. These authors propose that these techniques be used to answer all statistical questions. This is a well written book that will help you understand the foundations of statistics, and will teach you a versatile approach.
StatSoft's Electronic Statistics Text Really more of a small encylopedia than a text. There is no attempt at organization or flow. Rather there are a bunch of articles on various topics, some quite good. A good place to look up statistical topics -- not a good place to get started with the basics.
VassarStats Online text of statistical basics, accompanied by a set of online calculators. A fairly conventional approach, from a behavioral sciences point-of-view, but very clear.
Engineering Statistics This book is very different than the others. Engineers don't approach statistics as a way to analyze experiments, but rather as a tool to help optimize processes -- a very different approach to the subject. If you are involved with manufacturing or quality control, this approach might be much more useful to you than the standard one.
The Little Handbook of Statistical Practice. by Gerard E. Dallal. Clear, practical discussions of how to interpret statistical results. Especially good for multiple regression and anova.
Mutivariate Statistics. Concepts, Models and Applications. by David Stockburger. The place to go to learn about multiple regression, factor analysis, cluster analysis, discrimant analysis and much more.
CK-12 Statistics. Teaches very basic ideas of probability and data presentation to middle and high school students. Includes Khan academy videos.
Open Intro. A full conventional well-written statistics text book with lots of exercises. It reads exactly as a text, and in fact you can buy a printed version. Lots of examples. Emphasizes confidence intervals. Nothing about multiple comparisons, survival analysis, nononlinear regression, or Bayes.
Collaborative statistics. Step-by-step guide, with lots of exercises, to doing statistical calcualations. Emphasizes how to do the calculations over how to interpet the resuts. Nothing about multiple comparisons, survival analysis, nononlinear regression, or Bayes.