Start Date

14-12-2012 12:00 AM

Description

Covariance-based structural equation modeling is a popular statistical technique in information systems research, providing a stringent test of model fit and allowing researchers to test multiple hypotheses in the same model. Structural regressions in such models are often assumed to represent the causal nature of the underlying reality as expressed by theory. The validity of conclusions drawn from covariance-based analysis is, however, challenged when models can be constructed that fit the observed covariances equally well as the tested model, but which have a different structure, expressing different underlying causal relationships. This research shows that a large proportion of studies in IS exhibit this issue. The dangers posed by covariance-equivalent models are highlighted using an example in the published literature, and recommendations are provided to IS researchers to address the problem.

Share

COinS
 
Dec 14th, 12:00 AM

A Note of Caution on Covariance-Equivalent Models in Information Systems

Covariance-based structural equation modeling is a popular statistical technique in information systems research, providing a stringent test of model fit and allowing researchers to test multiple hypotheses in the same model. Structural regressions in such models are often assumed to represent the causal nature of the underlying reality as expressed by theory. The validity of conclusions drawn from covariance-based analysis is, however, challenged when models can be constructed that fit the observed covariances equally well as the tested model, but which have a different structure, expressing different underlying causal relationships. This research shows that a large proportion of studies in IS exhibit this issue. The dangers posed by covariance-equivalent models are highlighted using an example in the published literature, and recommendations are provided to IS researchers to address the problem.