Dynamic Modeling with Intensive Longitudinal Data: One-Step and Two-Step DSEM Approaches

Area of Science:

• Psychometrics
• Quantitative Psychology
• Statistical Modeling

Background:

• Intensive longitudinal data (ILD) analysis requires sophisticated statistical methods.
• Dynamic structural equation modeling (DSEM) is a powerful technique for ILD.
• Comparing one-step and two-step DSEM approaches is crucial for accurate analysis.

Purpose of the Study:

• To evaluate and compare the performance of one-step and two-step DSEM for ILD.
• To investigate the impact of auxiliary variables on two-step DSEM.
• To provide recommendations for DSEM application in ILD research.

Main Methods:

• A simulation study was conducted to compare DSEM approaches.
• One-step DSEM estimates within- and between-person models simultaneously.
• Two-step DSEM separates within-person and between-person model estimation, with and without auxiliary variables.

Main Results:

• Two-step DSEM without auxiliary variables demonstrated estimation bias, low coverage, and deflated Type I error rates.
• One-step DSEM and two-step DSEM with auxiliary variables performed satisfactorily with sufficient data (≥30 time points, ≥100 individuals).
• Auxiliary variables improved the performance of two-step DSEM.

Conclusions:

• One-step DSEM is a reliable approach for analyzing ILD.
• Two-step DSEM requires careful implementation, preferably with auxiliary variables, for valid results.
• Researchers should consider these findings when choosing DSEM methods for ILD analysis.