general

I taught prob/stat from 99-02. The first year we used Tanis and Hogg 5th, the second 6th Edition, and the last two semesters were the current Larsen and Marx (much nicer IMO, at least on the readability and organization factor). I felt the 6th was a step backwards from the 5th in terms of organization and the 5th wasn't great for that. In fact, the 6th felt a lot more like the old 3rd edition I had when I was taking prob/stat for the very first time. In my experience, many of the flaws that other reviewers have noted are present in this book:

-Confusing transitions between theorems and examples in the main text, with theorems and examples not clearly delineated from surrounding text (particularly where they end), making it hard to use the book for reference as you work problems;
-Examples that have too many steps skipped;
-Problems which are often much more difficult than the examples in the text (that said, many of the problems in T&H are quite good, which I suspect explains the longevity of it as a text, well that, simple inertia among profs who have their lectures already written, and the publisher's sales staff :).

*I* could use it as a reference as it does cover a lot of topics pretty well. So what. If I'm teaching the material I *better* have a mastery over the material in the book! Student have got to be able to use it and they don't have that mastery. Bad organization simply obscures the material which, while not difficult from a straight mathematical standpoint, IS pretty difficulty conceptually IME.

I feel much better after reading other reviews of this book - I am not the only one or the reason for this . This book is very deficient . . Especially for undergraduate learners .

While studying statistics with this book I thougth I lack a gene or something because I couldn't make much sense of the subject. Today I know that the book is conceptually wrong.

The book is basically stuck in the 1930's, ignoring major breakthroughs by Shannon, Kolmogorov and others. Fundamental concepts like entropy, individual randomness and typicallity are completely missing.

Furthermore, the book is heavily biased towards hypothesis testing, with little mentioning of Bayesian statistics.

From a pedagogic point of view it must be said that statistical concepts emerged historically and are better understood when placed in the context of the original problems. This book does not do that at all.