Structural Equation Modeling With Ordinal Indicators: A Comparison of Three Estimation Methods

• In Event: Estimation and Interpretation of Educational Data

Abstract

This study aims to further investigate the question of “can ordinal indicators be treated as continuous in SEM?”. We propose to conduct a Monte Carlo study to examine and compare the performance of ML, robust ML, DWLS, and ULS methods in analyzing SEM with ordinal indicators containing at least five response categories. The manipulated factors are model structure, number of response categories, distribution of ordinal indicators, similarity of distribution across ordinal indicators, and sample size. Results from preliminary analysis indicate that no one method outperform the other methods in terms of parameter and standard error estimates. All methods over-reject the true model based on chi-square test statistic, but rejection rates from ML methods are relatively small.

Authors

• Lu Liu, Florida State University

• Yanyun Yang, Florida State University