Structural Equation Modeling as a Global Modeling Framework with Examples from the Health Sciences

• Activity Number: 667
• Type: Invited
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Social Statistics Section
• Author(s): Sharon L. Christ*
• Companies: Purdue University
• Keywords: structural equation modeling ; generalized linear modeling ; systems ; health

Abstract

Structural equation modeling (SEM) is a multivariate modeling framework that subsumed many modeling and estimations options. SEM allows for the simultaneous estimation of systems of equations where variables may have different roles in different equations. Because of the ability to model multiple processes together, more "holistic" models may be specified and estimated. Another advantage of SEM is the ability to model constructs that are not directly observable using latent variables and to incorporate these variables into larger models. A third advantage of SEM is the ability to test complex linear and non-linear combinations of parameter estimates from within and across equations. Also, a recent extension to classical SEM, which assumed approximately normally distributed, continuous outcomes is generalized linear modeling capabilities. Now outcomes with different parametric and non-parametric distributions may be included in the same model. This is an important extension for the health sciences where event time, binary, and Poisson outcomes are common. The primary advantages and some current limitations of SEM will be demonstrated using examples from health research applications.