ABSTRACT
Structural equation modeling (SEM) represents a theory-driven data analytical approach for the evaluation of a priori specified hypotheses about causal relations among measured and/or latent variables. Such hypotheses may be expressed in a variety of forms, with the most common being measured variable path analysis (MVPA) models, confirmatory factor analysis (CFA) models, and latent variable path analysis (LVPA) models. For analyzing models of these as well as more complex types, SEM is not viewed as a mere statistical technique but rather as an analytical process involving model conceptualization, parameter identification and estimation, data-model fit assessment, and potential model re-specification. Ultimately, this process allows for the assessment of fit between (typically) correlational data, obtained from experimental or non-experimental research, and one or more competing causal theories specified a priori; most common SEM applications are not designed for exploratory purposes. Software packages such as AMOS, EQS, lavaan, LISREL, Mx, and Mplus are utilized to complete the computational, but not the substantive aspects of the overall SEM process. For contemporary treatments of SEM we recommend texts by Byrne (1998, 2006, 2012, 2016), Kline (2016), and Loehlin and Beaujean (2017), or, for more advanced readers, books by Bollen (1989), Hancock and Mueller (2013), Hoyle (2012), and Kaplan (2008). Specific desiderata for applied studies that utilize SEM are presented in Table 33.1 and explicated subsequently.
