Testing non-nested structural equation models

Abstract

In this paper, we apply Vuong's (1989) likelihood ratio tests of non-nested models to the comparison of non-nested structural equation models. Similar tests have been previously applied in SEM contexts (especially to mixture models), though the non-standard output required to conduct the tests has limited their previous use and study. We review the theory underlying the tests and show how they can be used to construct interval estimates for differences in non-nested information criteria. Through both simulation and application, we then study the tests' performance in non-mixture SEMs and describe their general implementation via free R packages. The tests offer researchers a useful tool for non-nested SEM comparison, with barriers to test implementation now removed.