The SAGE Encyclopedia of Communication Research Methods

Abstract

Structural equation modeling (SEM) is a versatile tool for conducting a wide range of multivariate statistical analyses, including multiple regression, mediation analysis, moderation analysis, and analyses of variance and covariance. Two specialized uses of SEM that appear frequently in communication research are confirmatory factor analysis (CFA) and path analysis. Confirmatory factor analysis specifies one or more unobserved constructs, or latent factors, that a number of observed indicators define. This analysis is useful for validating the composition of multiple-item indices or scales. The other common use of SEM, path analysis, estimates correlation and regression paths among structural nodes, which may include both observed variables and latent factors. When path analysis includes latent factors, the definition of those factors is equivalent to CFA and is the basis of a measurement model. The specification of paths among latent factors and observed variables constitutes a structural model. Whatever the intended use of SEM in communication research, it should be based on careful theoretical considerations. This entry conducts a detailed examination of the general approaches researchers can take to utilize SEM, as well as its various uses.