Yesterday I picked up Statistics in Plain English, by Timothy C. Urdan, and I tell ya I can't put it down. From my review:
I think I can say, without fear of hyperbole, that this is the best math book in the history of the entire universe. The fact that there are only six reviews of the book so far, instead of six hundred, hints at the fundamental problem I personally see in math education: it looks harder than it is because we communicate so poorly about it. Urdan communicates clearly and naturally, so the chilly math textbook mystique drops away, and you are left with a functional vocabulary of basic stats techniques.
Urdan starts with the assumption that all humans can understand and benefit from statistical techniques. By assuming that, he makes it true. He not only defines every term and every symbol he uses -- which is already amazing -- but the new terms and definitions are summarized at the end of each chapter. He lays out lots of context and many straightforward and interesting examples. The chapters are short, which gives you a nice feeling of accomplishment and plenty of breaks to think. He even humanizes the experience by speaking in the first person, expressing personal preferences, and even cracking the occasional joke. It's like talking about math over tea with a good friend.
In the modern data space, there's a great shortage of people who have a comfortable intuition for stats. If this book were in every undergraduate class, I'd wager that shortage would just go away.