Latent Interaction Modeling Using Parcels: Comparing Product Indicator and Latent Moderated Structural Equation Methods

• In Event: Statistical Challenges

Abstract

This study evaluates two latent interaction modeling methods when parcels are used as measurement indicators. A Monte Carlo study is designed to compare performance of UPI and LMS under different parceling strategies and data conditions where items are ordinal. The manipulated factors are interaction effect size, factor loadings, number of response categories, combination of item distributions across factors, and sample size. A preliminary analysis is conducted. Solutions from all replications are admissible. Results suggest that LMS performs well in estimating latent interaction effects. UPI performs slightly worse in terms of parameter estimates and coverage rates than LMS.

Authors

• Lu Liu, Florida State University
  Presenting Author

• Qian Zhang, Florida State University
  Non-Presenting Author

• Yanyun Yang, Florida State University
  Non-Presenting Author

• Wenyi Li, Florida State University
  Presenting Author