A First Course in Design and Analysis of Experiments

Rating:
Summary: My favorite design book
Review: I have taught from this text at the senior/applied masters level three times, and my enthusiasm increases as time goes on. Students generally share my favorable opinion. I have also taught using other texts (incuding Montgomery), and this one is my current favorite.
Oehlert takes design seriously, not just the analysis of experiments. He is obviously an experienced statistician with deep knowledge of both practice and theory. As a bonus, he writes well and uses excellent real examples.

Aspects of the text that stand out as different from other texts include:
1. A detailed explanation of different error rates for multiple comparisons. More detail is present than is useful for a beginning student, but the exposition is excellent as a reference for those who need it.
2. An unusually well-informed (and practical) discussion of assumptions.
3. Discussion of SAS Type II errors. Common practice is not always the most sensible, and the author's advocacy of the Type II approach for many problems is compelling.
4. A good discussion of mixed model assumptions (restricted and unrestricted models). I have not seen a comparable exposition of this potentially confusing issue.
5. The use of Hasse diagrams for mixed models. I had not worked with Hasse diagrams before I used this text, but find them to be useful tools for analysis of complex designs.

My only quibble is that some items in chapter 13 could be introduced earlier....although probably not fully covered. In particular, RCBDs without interactions could appear (with appropriate caveats) along with factorial designs. I confess some ambivalence on this issue, noting that I only quarrel because I am starting to get rushed for time by chapter 13.

If you want a cookbook, go elsewhere. If you want a highly mathematical approach, this is also not for you. For a serious treatment of real statistical issues, however, both analysis and design, I doubt if you can do much better.

Rating:
Summary: My favorite design book
Review: I have taught from this text at the senior/applied masters level three times, and my enthusiasm increases as time goes on. Students generally share my favorable opinion. I have also taught using other texts (incuding Montgomery), and this one is my current favorite.
Oehlert takes design seriously, not just the analysis of experiments. He is obviously an experienced statistician with deep knowledge of both practice and theory. As a bonus, he writes well and uses excellent real examples.

Aspects of the text that stand out as different from other texts include:
1. A detailed explanation of different error rates for multiple comparisons. More detail is present than is useful for a beginning student, but the exposition is excellent as a reference for those who need it.
2. An unusually well-informed (and practical) discussion of assumptions.
3. Discussion of SAS Type II errors. Common practice is not always the most sensible, and the author's advocacy of the Type II approach for many problems is compelling.
4. A good discussion of mixed model assumptions (restricted and unrestricted models). I have not seen a comparable exposition of this potentially confusing issue.
5. The use of Hasse diagrams for mixed models. I had not worked with Hasse diagrams before I used this text, but find them to be useful tools for analysis of complex designs.

My only quibble is that some items in chapter 13 could be introduced earlier....although probably not fully covered. In particular, RCBDs without interactions could appear (with appropriate caveats) along with factorial designs. I confess some ambivalence on this issue, noting that I only quarrel because I am starting to get rushed for time by chapter 13.

If you want a cookbook, go elsewhere. If you want a highly mathematical approach, this is also not for you. For a serious treatment of real statistical issues, however, both analysis and design, I doubt if you can do much better.

Rating:
Summary: One of the few truly modern MS level texts
Review: Unlike the market-leaders that are showing their age, this book takes a modern point of view of experimental design, not excessively tilted towards industrial settings. It's also the only book showing how to use Hasse diagrams to find expected mean squares, which is by far the easiest way. The book also pays considerable attention to the design, not just the analysis, of experiments. It emphasizes practical, rather than mathematical, depth and insight. It's terrific.

Rating:
Summary: Not very good
Review: We are using this book for a Masters level graduate course in Experimental Design. The book is poorly written. It seems like it is basically a compilation of lecture notes. I would use Montgomery over this text. Even our professor admits that it is useless to study from.

The plus points are that it covers certain things that other text books don't (Error Rates, SNK, etc.). These are only covered minimally, though, and don't make up for the poor coverage of most of the other subjects in the book.