Abstract
Principles and Practice of Structural Equation Modeling, 4th editionRex B. KlineNew York: The Guilford Press 2016ISBN 978-1-4625-2334-4Softcover, US$65, 534 pp.Kline’s fourth edition is reasonably strong but improvable. The text aims to introduce newcomers to fundamental structural equation modeling (SEM) principles, but tends to confuse “Principles” with “Rules.” Rules having insufficient grounding in principles leave readers ill-prepared for understanding and responding to changes in previously traditional “rules”—such as those concerning model testing, and latents having single indicators. SEM’s foundations would be clearer if Kline began by presenting structural equation models as striving to represent causal effects—a commitment that differentiates structural equation models from regression and encourages model testing. I begin this review by summarizing the covariance/correlation implications of three simple causal structures, which pinpoints multiple text improvements and underpins the discussions of measurement and model testing that follow. Causal structuring also grounds my later comments regarding modelling means/intercepts and interactions. A file of Supplementary Sections expands on several points and lists multiple editorial corrections you might pencil into your copy of Kline’s text.
Highlights
Kline’s fourth edition is reasonably strong but improvable
Kline’s emphasis on multiple factor-structured indicators results in insufficient attention to measurement error variance in routinely used variables like age, sex, and education, which rarely have more than a single indicator
This double-scaling forces their Job Satisfaction latent to contribute exactly 1.0 unit to the variance of their Job Satisfaction indicator, when that indicator has a variance of 0.9392 = .88. This forces a model-data inconsistency, explaining why Kline’s model fits better than their model, and should have produced an impossible negative error variance estimate for Houghton and Jinkerson’s scaling indicator. Kline missed this opportunity to show how model misspecification can lead to problematic estimates, and missed the opportunity to caution that the literature contains enough errors to warrant routine caution and checking
Summary
Kline’s fourth edition is reasonably strong but improvable. The text aims to introduce newcomers to fundamental structural equation modeling (SEM) principles, but tends to confuse “Principles” with “Rules.” Rules having insufficient grounding in principles leave readers ill-prepared for understanding and responding to changes in previously traditional “rules”—such as those concerning model testing, and latents having single indicators. The causal world makes variations in one variable (X ) produce coordination, correlation, or covariance (Cov (XY )) between the causal variable and the effect. The partitioning of the causal world partitions the variance in the effect (Y ) but with the wrinkle that a portion of Y’s variance comes from coordination/covariance between the values of the causes, not merely from variations in the values of those causes. This variance equation is fundamental to: understanding why some explained variance cannot be uniquely attached to a specific cause,. A called a disturbance or error variable because it was not observed)
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call