Evaluating methods for handling missing ordinal data in structural equation modeling

• Main Authors: Jia, F. (Author), Wu, W. (Author)
• Format: Article
• Language: English
• Published: Springer New York LLC 2019
• Subjects: article; human; latent class analysis; Latent Class Analysis; Missing ordinal data; Monte Carlo method; Monte Carlo Method; Multiple imputation; Robust estimation; scientist; structural equation modeling; Structural equation modeling
• Online Access: View Fulltext in Publisher

 Missing ordinal data are common in studies using structural equation modeling (SEM). Although several methods for dealing with missing ordinal data have been available, these methods often have not been systematically evaluated in SEM. In this study, we used Monte Carlo simulation to evaluate and compare five existing methods, including one direct robust estimation method and four multiple imputation methods, to deal with missing ordinal data. On the basis of the simulation results, we provide practical guidance to researchers in terms of the best way to deal with missing ordinal data in SEM. © 2019, The Psychonomic Society, Inc.