Abstract
Structural equation modeling (SEM) is a flexible and powerful statistical analysis framework that is appropriate for much of the data used by social work researchers (Bowen & Guo, 2012). By combining confirmatory factor analysis with path analysis, SEM permits (a) the sophisticated evaluation of social work scales and instruments and (b) the testing of hypotheses about complex developmental processes and intervention effects. In short, SEM can provide answers to a host of valuable social work research questions both cross-sectional and longitudinal in nature. However, SEM has seen relatively slow uptake among social work researchers for two key reasons. First, SEM is not always taught in social work education, putting these analyses beyond the core statistical training of many researchers. Second, until recently most of the common statistical software packages (e.g., SPSS, Stata) were not capable of SEM analyses, necessitating the use of specialized software and the commensurate expenditure of additional time and resources. In all, SEM has, despite its significant advantages and relative ease of use, been too often avoided by social work researchers.
The 2015 iteration of Kelloway’s Using Mplus for Structural Equation Modeling: A Researcher’s Guide continues on the agreeable goal of the first edition to provide a reader-friendly and approachable introduction to SEM using Mplus software. Those who are new to this field of analyses may wonder aloud why Mplus, in particular, makes an attractive choice for SEM. In short, Mplus is exemplary in four key ways that put it at or near the top of most SEM researchers’ choice in statistical software packages (Byrne, 2012). First, Mplus has an extensive ability to handle the categorical and nonnormal data most often found in social work research. Second, Mplus has flexibility regarding the testing of a broad range of models of various nested structures. Third, the Mplus user interface features a straightforward and relatively easy-to-learn syntax structure that requires a relatively minimal time investment in order to become proficient. Last, Mplus is greatly bolstered by excellent documentation and supplemental support from both an online community of users and the software developers themselves (see http://www.statmodel.com).
The book’s author, Dr. E. Kevin Kelloway, is currently Canada Research Chair in Occupational Health Psychology and Professor of Psychology at Saint Mary’s University in Halifax, Nova Scotia. Dr. Kelloway received his PhD in organizational psychology from Queen’s University in Kingston, Ontario, in 1991. He is an internationally recognized expert in occupational and organizational psychology, with over 150 published articles, book chapters, and technical reports, and in 14 authored/edited books. His particular areas of expertise relates to issues of leadership, occupational health psychology, and human resource management within both public and private enterprise. A great deal of Dr. Kelloway’s research has either directly used SEM analyses or contributed to the body of knowledge regarding the application of SEM methods.
Nine chapters comprise the book, progressing roughly from basic theoretical discussions of SEM through to basic SEM and Mplus usage and ending with more complicated applications. Chapters 1 through 3 provide a useful, software-agnostic introduction to the theory, development, and core constructs of SEM. Dr. Kelloway provides numerous reader-friendly figures, examples, and equations herein that will likely appeal to social work researchers. The material presented herein is clear, concise, and generally well presented throughout, though by no means should the reader rely solely on this content as a complete introduction to SEM. Chapter 4 provides an overview of the particulars of Mplus, including coverage of its file structure, syntax language, and analytic procedures. Chapters 5 through 7 are devoted to the common SEM-related tasks of confirmatory factor analysis, observed variable path analysis, and latent variable path analysis, respectively. The latter section, comprising what is most commonly known to the majority of researchers as “SEM,” is likely the section of the book in which many readers will likely have the most interest. Chapters 8 and 9 introduce more advanced SEM applications dealing with longitudinal and multilevel analyses, respectfully. The book is completed by an adequate list of references.
There is much to like about this book. In particular, the formatting is exemplary. Near-perfect facsimiles of actual Mplus files are used, with appropriate comments and explanations clearly delineated throughout in boldface text. Reading the book’s content and digesting important pieces of information is easy and enjoyable throughout. The book is also quite thorough, containing relevant information on a variety of analyses of interest to social work researchers. More generally, Dr. Kelloway’s expertise shines throughout the presentation of information. The recommendations given are fully aligned with current SEM best practices, making the content useful, relevant, and a great starting point for researchers seeking to jump into their own SEM analyses.
Limitations of the book are not major but may be of importance to some readers. First, although Dr. Kelloway makes use of the so-called “real-world” examples throughout the text, these examples are not expounded upon to full fruition and serve, mostly, as touchstones. Some readers will long for a start-to-finish SEM example analysis as can be found in some other books. Second, Dr. Kelloway makes only brief forays into two realms which are likely to be of particular interest for social work researchers: (a) analyses involving nonnormal, ordinal data and (b) analyses involving high levels of missing data. Each topic is given only a brief mention in the text and many readers will likely want for more information regarding these crucial topics. Third, the book lacks numerous potentially useful add-ons that could enhance its content. Content such as sources for further SEM study, a discussion of common Mplus error messages, helpful tips for conducting analyses, and others would have been appreciated. It should be noted that the book lacks a summative chapter—perhaps such content would be a helpful addition to a potential third edition of the text?
Overall, Dr. Kelloway’s Using Mplus for structural equation modeling: A researcher’s guide constitutes a welcome addition to the SEM research base and will likely find many ardent fans within social work. It is recommended for those who are either new to SEM and/or Mplus and can also serve as a helpful refresher to those seeking to bolster existing knowledge and skills. It is not intended to be a one-stop-shop resource for conducting SEM, but it is certainly a worthwhile text that deserves consideration.