Model comparison is certainly an improvement over the old-school zoo of tests, but this book still manages to over-complicate plenty. (It was the impetus for me writing Contrast codes are an implementation detail, for example.) It was a course text and my course notes cover all the actual content. This post is just to warn people off the book itself.
I see two reasonable approaches to writing a statistics textbook:
- Rigorously prove everything
- Focus on intuition and model-building
This book takes the worst of both. It leaves software tools to the appendix, perhaps assuming that we prefer to use a hand calculator as required by the exam. It drowns the reader in equations but never bothers to derive them, or any of the underlying theory. It’s a sort of cargo-cult dedication to rigor, where the more LaTeX we use the more rigorous the book is.