Best Practices in Structural Equation Modeling

Abstract

S tructural equation modeling (SEM) has evolved into a mature and popular methodology to investigate theory-derived structural/causal hypotheses. Indeed, with the continued development of SEM software packages such as AMOS (Arbuckle, 2007), EQS (Bentler, 2006), LISREL (Joreskog & Sorbom, 2006), and Mplus (Muthen & Muthen, 2006), SEM “has become the preeminent multivariate method of data analysis” (Hershberger, 2003, pp. 43–44). Yet, we believe that many practitioners still have little, if any, formal SEM background, potentially leading to misapplications and publications of questionable utility. Drawing on our own experiences as authors and reviewers of SEM studies, as well as on existing guides for reporting SEM results (e.g., Boomsma, 2000; Hoyle & Panter, 1995; McDonald & Ho, 2002), we offer a collection of best practices guidelines to those analysts and authors who contemplate using SEM to help answer their substantive research questions. Throughout, we assume that readers have at least some familiarity with the goals and language of SEM as covered in any introductory textbook (e.g., Byrne, 1998, 2001, 2006; Kline, 2005; Loehlin, 2004; Mueller, 1996; Schumacker & Lomax, 2004). For those desiring even more in-depth or advanced knowledge, we recommend Bollen (1989), Kaplan (2000), or Hancock and Mueller (2006).