Abstract
The purpose of this study was to evaluate the sensitivity of selected fit index statistics in determining model fit in structural equation modeling (SEM). The results indicated a large dependency on correlation magnitude of the input correlation matrix, with mixed results when the correlation magnitudes were low and a primary indication of good model fit. This was due to the default SEM method of Maximum Likelihood that assumes unstandardized correlation values. However, this warning is not well-known, and is only obscurely mentioned in some textbooks. Many SEM computer software programs do not give appropriate error indications that the results are unsubstantiated when standardized correlation values are provided.
DOI
10.22237/jmasm/1493597040
Recommended Citation
Rose, S. A., Markman, B., & Sawilowsky, S. (2017). Limitations in the systematic analysis of structural equation model fit indices. Journal of Modern Applied Statistical Methods, 16(1), 69-85. doi: 10.22237/jmasm/1493597040
Included in
Applied Statistics Commons, Social and Behavioral Sciences Commons, Statistical Theory Commons
