Path Analysis With Categorical Outcome Data: A Closer Look of the Nonnormality Effects

• In Event: Advanced Structural Equation Modeling and Hierarchical Modeling at Roundtable

Abstract

In empirical studies, we often see researchers conducted the path analysis with a mixed continuous and categorical data, especially using continuous variables to predict a categorical outcome. In our Monte Carlo study, we compared the functions of several estimators under various conditions when the path model had mixed continuous and categorical data. The estimator included in this study were robust maximum likelihood (MLR-probit and MLR-logit) and weighted and unweighted least square adjusted by mean and variance (WLSMV and ULSMV). Preliminary results were illustrated in overall model fit, item level parameter estimates, power of path coefficients, and fit indices in various conditions. Difference between estimators are expected to see and suggestions will be made after full analysis completed.

Authors

• Ji Li, University of Arkansas for Medical Sciences

• Wen-Juo Lo, University of Arkansas

• Xinya Liang, University of Arkansas

• Chunhua Cao, The University of Alabama

• Paula Castro Abbott, University of Arkansas at Fayetteville

• Nnamdi Chika Ezike, University of Arkansas at Fayetteville