Examining Cross-Lagged Effects Using Corrected Composite Scores Under Longitudinal Measurement Non-Invariance

• In Event: Methodological Innovations in Structural Equation Modeling

Abstract

This study examines cross-lagged effects using composite scores of items measuring the latent variables while correcting for measurement errors. Focusing on scenarios where longitudinal measurement non-invariance (MNI) results in varying composite score reliabilities over time, this study considers three correction approaches: one uses time-invariant reliability for correction assuming longitudinal measurement invariance whereas the other two account for longitudinal MNI. A Monte Carlo study is designed to evaluate the three approaches under small-sample conditions. The manipulated factors include cross-lagged effects, composite score reliability at initial occasion, number of latent variables involving MNI, number of non-invariant items, non-invariance magnitude, and sample size. Preliminary results revealed promising performance of three approaches. They showed trivial estimation bias and generally well-controlled type I error rates.

Authors

• Lu Liu, Florida State University

• Yanyun Yang, Florida State University