Structural Equation Models: From Paths to Networks (Westland 2019)

Book review Westland’s (2019) Structural Equation Models: From Paths to Networks offers a concise, well-written, and non-technical reference for SEM. The textbook comprises 149 pages structured into eight chapters. The chapters are largely independent of one another, allowing them to be easily covered in a different order. The content of Westland’s (2019) textbook makes it very attractive as additional reading for methodological courses devoted to factor-based SEM and wanting to delve into the method’s historic roots. Readers will also appreciate the various overview tables offering excellent summaries of selected contents, as well as the book’s index, which allows them to quickly identify topics and key terms of interest. Whereas most textbooks focus on either factor-based or component-based SEM (e.g., Hair et al. 2017; Kline 2016), Westland’s (2019) book is unique in that it showcases the full range of SEM methodologies, starting with Wright’s (1921) path analysis, followed by partial least squares (PLS) path modeling (e.g., the difference between it and PLS regression is explicitly addressed), full-information covariance-based SEM, and recent neural network-based approaches. By placing these methods in a historical context, the book provides insights into their similarities and differences (e.g., the Chicago School and the Scandinavian School), which certainly explain why some disciplines have clustered around one or the other SEM method. This book has many other valuable aspects. Most notably, Westland (2019) strongly emphasizes research design issues, which often receive scant attention in textbooks. These issues have a fundamental bearing on the implications drawn from any SEM analysis (Rigdon et al. 2020). For example, Chapters 4 and 5 offer a detailed account of data adequacy in SEM, drawing specific attention to Likert scale data. Applied researchers will particularly benefit from the various guidelines for data collection and preparation, which go beyond the standard power analyses covered in popular textbooks. For example, Westland (2019, p. 101) derives a metric for computing sample sizes required to offset information loss through the use of Likert scales rather than continuous metrics. The explications in Chapter 7 on research protocols are particularly noteworthy. Common topics, such as the nature of models, theory building, and hypothesis testing, are presented concisely to offer scope for subjects one would not expect in a compact reference guide. For example, Westland (2019) issues cautionary warnings about behavioral biases, such as apriorisms, that impact causal inference. Given increasing concerns about reproducibility (Nuzzo 2015), users of SEM should take these warnings seriously, as causal inferential conclusions are only supported under specific conditions (Bollen and Pearl 2013)