am 13. November 2013
In the past 10 years or so, Andrew Hayes has established himself as one of the leading researchers in the statistical analysis of mediation and moderation. What makes his work unique is his understanding of the everyday problems that non-methodologists are confronted with in their work. His PROCESS macro, which has facilitated and improved the analysis of mediation and moderated mediation patterns, clearly attests to the service-oriented character of his work.
In "Introduction to Mediation, Moderation, and Conditional Process Analysis," Hayes continues with his service-oriented work for social scientists who are interested in patterns of how variables are related to, or influence, each other. The book covers much of the issues that Hayes has dealt with in the past decade, notably his work on mediation, the proper investigation of moderation, and the combination of the two, moderated mediation, that is. Similar to Hayes' previous work, the book is largely written in a non-technical manner. His target audience are researchers who aim at rigorous analysis of their data, but may not have the ability or time to work through the more technical statistical literature. Hayes does not assume too much statistical knowledge from his readers, which makes the book easily accessible. Actually, in chapters 1 to 3, Hayes explains the basics of correlation and causality, as well as simple and multiple linear regression. Still, the book is not suitable for statistical novices in my view. A previous rough understanding of multiple regression analysis and a basic grasp of traditional (stepwise, "Baron and Kenny" type) approaches to mediation are helpful to take full advantage of the book.
The book is divided in four parts. In part 1, Hayes outlines fundamental statistical concepts, such as correlation and regression analysis. Readers who do not deal with such issues in their daily work should definitely not skip this part. Part 2 deals with mediation analysis. This part not only covers simple mediation analysis, but also models with multiple mediators, and answers a lot of question that researchers encounter in their daily research: What do my effect sizes really mean? Shall I choose structural equation modeling or regression? And how do I analyze models with multiple independent or dependent variables? In part 3, Hayes deals with moderation analysis and issues, such as the post-hoc probing of interaction effects, interactions in which the moderating variable is dichotomous or quantitative, and cases with more than one moderator. Part 4, finally, focuses on conditional process analysis. This part, undoubtedly the most demanding, is probably the one where also the more advanced reader finds a lot to learn. Hayes starts with example of conditional process models in the literature, outlines conditional process analysis with PROCESS on the basis of many helpful examples, and eventually also deals with more advanced issues, such as whether a variable can simultaneously mediate and moderate another variable's effect. In sum, the four parts cover a lot of ground and will leave the reader with a very sound, up-to-date knowledge of what the rigorous analysis of mediation, moderation, and conditional processes means.
The book uses many examples from the social sciences. The use of the PROCESS macro is outlined in detail, including an appendix in which Hayes walks the reader step by step through the macro. Formulas are kept to a minimum and, if used, relatively easy to understand on the basis of knowledge from high-school mathematics. Hayes also devotes much attention to how to report results from the various analyses in research papers, doubtlessly an invaluable help for those who struggle with the statistical presentation of their results. If there is something that can be improved about the book, then it may be the presentation of the outputs. A simple overview that explains *in the output* what the various parameters mean may have been helpful, certainly for those who encounter such an ouput for the first time.
In sum, I believe that Hayes' book is a thoroughly needed, well-written and easily accessible on one of the key issues in the social sciences. You can use the book in teaching more advanced courses on statistics, for improving your statistical skills, or simply as a source for various questions that may arise in your mediation and moderation analyses. Either way, you will have a high-quality companion that will serve you nicely. In my view, this is currently the best book available on mediation, moderation, and the combination of the two.