Empirical and conceptual problems with longitudinal trait-state models: introducing a trait-state-occasion model

Area of Science:

• Psychometrics
• Structural Equation Modeling
• Psychological Measurement

Background:

• Latent trait-state-error (TSE) and latent state-trait autoregression (LST-AR) models are advanced methods for longitudinal psychological construct analysis.
• Previous applications of TSE and LST-AR models have been hindered by empirical and conceptual challenges.

Purpose of the Study:

• To evaluate the empirical limitations of the TSE and LST-AR models.
• To identify a more robust and broadly applicable structural equation model for longitudinal psychological research.

Main Methods:

• Monte Carlo simulation to assess TSE model performance under various conditions (sample size, number of waves, occasion factor stability).
• Mathematical analysis to determine the applicability constraints of the LST-AR model.
• Comparative analysis of the trait-state-occasion model against TSE and LST-AR.

Main Results:

• TSE models produced improper solutions with small sample sizes, few waves, or extreme occasion factor stability.
• LST-AR models are limited to constructs with increasing autocorrelation over time.
• The trait-state-occasion model demonstrated fewer empirical issues and broader applicability.

Conclusions:

• The TSE model requires careful consideration of sample size, number of measurement occasions, and parameter values to avoid improper solutions.
• The LST-AR model is suitable only for specific types of longitudinal data.
• The trait-state-occasion model presents a more reliable and versatile option for examining longitudinal changes in psychological constructs.