Structural Equation Methodology

Abstract

AbstractSince the 1970s, structural equation modeling (SEM) has been applied for the analysis of causal relationships in economic and social sciences (Götz and Liehr-Gobbers 2004: 714; Henseler 2005: 70; Herrmann et al. 2006: 35). SEM is a multivariate statistical technique that – by integrating different regression- and factor-analytic methods – allows the testing and estimation of theoretically derived, casual relationships between variables (Bortz 1993: 436; Rigdon 1998: 251). Its high popularity in economic and social sciences is mainly attributed to two factors. First, it is the only multivariate statistical method that allows one to simultaneously assess the quality of construct measurement in terms of reliability and validity on the one hand, while on the other hand estimating the strength of a relationship between constructs (Backhaus et al. 2008: 511; Henseler 2005: 70). Second, SEM enables scientists to measure not only observable (manifest) variables, but also unobservable (latent) variables (Chin 1998b: 296; Chin and Newsted 1999: 307; Herrmann et al. 2006: 35; Rigdon 1998: 251). As latent variables are very common in economic and social sciences, the ability to model latent variables is of particular importance (Chin 1998b: 296). Due to the fact that latent variables cannot be observed directly, they are assigned manifest variables – which can be derived empirically and measured on the basis of metric scales – in a measurement model (or outer model) (Backhaus et al. 2008: 513; Henseler 2005: 70), as illustrated in Fig. 3.1KeywordsLatent VariableStructural Equation ModelingMeasurement ModelLatent ConstructFormative IndicatorThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.