Masking Misfit in Confirmatory Factor Analysis by Increasing Unique Variances: A Cautionary Note on the Usefulness of Cutoff Values of Fit Indices
This article investigates the fit indices use in structural equation modeling. “Cut-off” values of fit indices were commonly used in modern psychology work where SEM is applied to determine the degree of the fit between the data and a model of concern. This paper particularly worked on the most commonly used chi-square test statistic fit index root-mean-square. The current study has a number of implications especially for practice: first, it has clearly shown the cut-off rule should be carefully used in consideration of the size of unique variance of the complexity of the model under test. However, this does not mean that approximate fit indices should be used because it have been seen that it over-estimate the goodness of fit. The article seems to suggest that there are fewer cases where SEM is applied and models evaluated correctly than the current cases in the field of psychology or social science.