Analyzing latent state-trait and multiple-indicator latent growth curve models as multilevel structural equation...

Area of Science:

• Psychometrics
• Longitudinal Data Analysis
• Statistical Modeling

Background:

• Latent state-trait (LST) and latent growth curve (LGC) models are common for longitudinal data.
• Standard single-indicator LGC models are analyzed using structural equation modeling (SEM) or multilevel (ML) frameworks.
• LST and multivariate LGC models, using multiple indicators, can also be specified within the ML framework.

Purpose of the Study:

• To demonstrate that multilevel SEM (ML-SEM) can be practically applied to LST and multivariate LGC models.
• To highlight the advantages of ML-SEM over single-level SEM (SL-SEM) for specific longitudinal data characteristics.
• To discuss the benefits and drawbacks of using ML-SEM for these complex models.

Main Methods:

• The study proposes specifying LST and multivariate LGC models within a multilevel SEM (ML-SEM) framework.
• The ML-SEM approach is compared to the traditional single-level SEM (SL-SEM) framework.
• An application using Mplus software is presented with ecological momentary assessment data from 158 youths.

Main Results:

• ML-SEM is practical for LST and LGC models with a large number of time points.
• ML-SEM accommodates individually-varying observation times, unequally spaced intervals, and incomplete data.
• The study provides an empirical example illustrating the application and considerations of ML-SEM.

Conclusions:

• ML-SEM offers a flexible and practical approach for analyzing complex longitudinal data with LST and LGC models.
• Researchers should consider ML-SEM when dealing with longitudinal datasets characterized by numerous observations, irregular time intervals, or missing data.
• While offering advantages, potential limitations of the ML-SEM approach should also be carefully evaluated.