Customer Reviews

Multiple Regression: Testing and Interpreting Interactions

5 star:		(4)
4 star:		(2)
3 star:		(0)
2 star:		(1)
1 star:		(0)

 

 

This book has revolutionized the way psychologists think about interactions. It provides step-by-step instructions on how to probe the moderating effects after you find a significant interaction in a multiple regression.

The basic idea about interaction is that the relationship between two variables were different according to a third variable. For example, some risk factors (such as poor family income) may affect children's academic achievement in a negative way. However, if the parents provide enough support on their children's study, then it's possible that the risk factors will no longer influence their children test scores. Therefore, with low support, risk factors are very effective, but with high support, risk factors have not effects. This book teaches you how to probe this relationship in a systematic way, it covers 2-way, 3-way interactions and also quardratic relationships.

If you fully understand this book, the techniques you have will be enough for a masters thesis in your area.

This book takes a very practical approach to the analysis of interactions in regression. No other book I've used has covered these topics as clearly or in as much depth. The extensive discussion of decomposing interactions is a prime example. With the push to replace old techniques of dichotomizing continuous variables with a continuous (regression) treatment of these variables (especially in psychology), this book is extremely important.

Back when the book was first published, I was completing doctoral research. Aiken & West provided the explanations and instructions that enabled me to complete my dissertation. Nowhere else have I seen the information they provide; seldom have I seen statistical treatments as clearly and easily explained. Like many in the social sciences, math was not my greatest intellectual ability. This book made computing and understanding regression interactions a relative breeze. One reviewer bemoaned the lack of information on interactions among categorical variables. I suppose he didn't read the preface that specifically explains the reason for the absence: such information is widely available in any good statistical text. What Aiken & West provide can't be found elsewhere in any real depth. I am ordering another copy of the book because I'm tired of loaning out my copy to colleagues, especially one who has now begun to copy whole chapters. Yes, it's that useful.

Yes, there is some good information and discussion in this book but for the price I would expect it to be more complete. For example, there is absoultely no mention of interactions between two categorical variables. I guess the authors ran out of steam. Also, the writing could have used some more refinement. I'd stick with Jaccard's volumes in the Sage Quantitative Applications series.

While the presentation and writing might not be quite perfect for some readers, this book provides the best coverage of handling interactions in multiple regression that I've yet seen. Everyone who does multiple regression with any regularity needs to have this book on their shelf.

If you want to get the most out of 2 continuous predictors in your data, then this book will be very useful. Excellent description of how to graph 2-way interactions in multiple regression equations, as well as theoretical recommendations for adding/removing higher order terms.

A colleague recommanded this book, when I was struggling with multip-effect interaction. The book proved to be an apt and excellent choice for this kind of problem. There are several chapters detailing the interactions, unlike many other regression textbooks, which tend to skim over this subject. If the authors could be more succint in their writing, this book would have definitely deserved a 5-star.