I have recently reviewed a large collection of DoE books (17), with the following requirements:
- Not a cookbook approach but geared towards understanding (hard requirement)
- Decently in-depth (hard requirement)
- Written with an understanding of the New Causal Revolution (nice-to-have)
- Uses Hasse diagrams to simplify understanding of variance structure (nice-to-have)
- Has problems or exercises, preferably with solutions, for self-study (must have problems, nice to have solutions)
- Does not use SAS (I hate SAS) - or at least has an alternative to SAS (hard requirement)
- Utilizes a depth-first approach rather than breadth-first (suits my learning style better - hard requirement)
- Geared towards upper-level undergraduate with prerequisites of mathematical statistics and linear models (hard requirement)
The books I reviewed were the following (only first author listed):
- Kaltenbach, Statistical Design and Analysis of Biological Experiments
- Casella, Statistical Design
- Montgomery, Design and Analysis of Experiments, 4th Ed.
- Montgomery, Design and Analysis of Experiments, 10th Ed.
- Montgomery, Design of Experiments: A Modern Approach
- Oehlert, A First Course in Design and Analysis of Experiments
- Lawson, Design and Analysis of Experiments with R
- Fisher, The Design of Experiments
- Morris, Design of Experiments: An Introduction Based on Linear Models
- Bailey, Design of Comparative Experiments
- Wu, Experiments: Planning, Analysis, and Optimization
- Dean, Design and Analysis of Experiments
- Box, Statistics for Experimenters
- Maxwell, Designing Experiments and Analyzing Data: A Model Comparison Perspective
- Mead, Statistical Principles for the Design of Experiments
- Cobb, Introduction to Design and Analysis of Experiments
- Kuehl, Design of Experiments: Statistical Principles of Research Design and Analysis, 2nd Ed.
I was unable to find any book that had all of the desired characteristics. Indeed, I have come to think that Requirement #3 is not satisfied anywhere.
Requirement #4 is only true for three books in the list (Oehlert, Kaltenbach, and Bailey).
Two books were too advanced (Casella and Morris), though Casella would likely be a terrific graduate-level text.
Very few books had answers to problems, and quite a few had no problems at all (Fisher and Kaltenbach, with Bailey and Mead having too few problems).
A number of books just presented a cookbook approach: here's the design and how you analyze it, with a very limited attempt at getting to the understanding (all the Montgomery books, Oehlert, Lawson, Kuehl). Maxwell was far too wordy, Cobb was too low-level and also very wordy. Wu was breadth-first.
I narrowed it down to two possibilities that looked promising: Dean and Box. I had to read the first few chapters of both to determine that Dean uses a depth-first approach, while Box has a breadth-first approach.
So I have landed on Dean, Voss, and Draguljic Design and Analysis of Experiments as my favored book for self-study, based on my requirements. It fails Requirements #3 and #4 and does not have solutions for problems. But it does well enough on the other requirements to be the best option.
Fisher is worth reading, but is not sufficient on its own to train you to be proficient, as there are no problems to work. He has very good explanations, and it would be a great supplement.