Modeling qualitative between-person heterogeneity in time series using latent class vector autoregressive models

Area of Science:

• Psychological Research
• Quantitative Psychology
• Time-Series Analysis

Background:

• Time-series data are crucial for understanding within-person dynamics in psychology.
• Vector autoregressive (VAR) models are commonly used to approximate these dynamics.
• Existing hierarchical models often assume quantitative heterogeneity across individuals.

Purpose of the Study:

• Introduce the latent class vector autoregressive (LC-VAR) model as an alternative to traditional methods.
• Address the lack of accessibility for LC-VAR models in applied psychological research.
• Provide practical tools and guidance for estimating and interpreting LC-VAR models.

Main Methods:

• Developed an accessible introduction to latent class VAR models.
• Conducted a simulation study to assess model estimation in realistic scenarios.
• Introduced the R package ClusterVAR for user-friendly LC-VAR model estimation.
• Provided a reproducible tutorial for modeling emotion dynamics using LC-VAR.

Main Results:

• The latent class VAR model effectively captures qualitative heterogeneity in within-person dynamics.
• Simulations demonstrated the feasibility of estimating LC-VAR models with applied data.
• The ClusterVAR package simplifies the application of LC-VAR models.
• The tutorial illustrates a complete workflow for LC-VAR analysis.

Conclusions:

• Latent class VAR models offer a valuable framework for understanding individual differences in psychological processes.
• The developed R package and tutorial enhance the accessibility and application of LC-VAR models.
• This approach advances the analysis of complex time-series data in psychological research.