The Performance of ML, GLS, and WLS Estimation in Structural Equation Modeling Under Conditions of Misspecification and Nonnormality

By U. Olsson, T. Foss, S. Troye, Roy D. Howell. 2000

682 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Management Science and Operations Research, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/706d7009511d34e692fd540f7475821f4d1b2ad7

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

In Management Science and Operations Research

In Statistics and Probability

Citations in the corpus

Structural Equations Modeling: Fit Indices, Sample Size, and Advanced Topics (2010, methodology): “Comparisons of estimation methods show maximum likelihood (ML) generally performs best, better than generalized least squares (GLS), and especially better than weighted least squares (WLS) (Ding, Velicer & Harlow, 1995; Olsson et al., 2000).”
Effects of Missing Data Methods in Structural Equation Modeling With Nonnormal Longitudinal Data (2009, background): “With complete data, some researchers ( Olsson, Foss, Troye, & Howell, 2000 ; Wang, Fan, & Willson, 1996 ) have suggested that when the sample size is over 2,500, the bias of the ADF estimation method is negligible and the method is appropriate.”
Parenting measures in the Longitudinal Study of Australian Children: construct validity and measurement quality, Waves 1 to 4 (2014, methodology): “If RMSEA is used with WLS then choosing a higher threshold is recommended (Olsson et al., 2000).”
Factor recovery by principal axis factoring and maximum likelihood factor analysis as a function of factor pattern and sample size (2012, background): “41 found that maximum likelihood was more stable and less biased in terms of both empirical fit and recovery of the underlying structure as compared with generalized least squares.”
Further Refinements in the Measurement of Exercise Imagery: The Exercise Imagery Inventory (2005, background): “In samples less than 5,000, violations of normality do not make a substantial difference in multivariate analysis generally and structural equation modeling in particular (Olsson et al., 2000).”
A Model of Event–Destination Image Transfer (2014, methodology): “After weighing the advantages and shortcomings of various estimation methods, the MLE method was finally adopted to estimate the model in the SEM analysis because of its robustness to the violation of the normality assumption (Olsson et al. 2000; Savalei 2008).”
A Monte Carlo Study of Recovery of Weak Factor Loadings in Confirmatory Factor Analysis (2006, result): “These findings are consistent with the studies of La Du and Tanaka (1989), Sugawara and MacCallum (1993), and Olsson et al. (2000).”