Timescale mismatch in intensive longitudinal data: Current issues and possible solutions based on dynamic structural...

Area of Science:

• Psychological Methods
• Quantitative Psychology
• Longitudinal Data Analysis

Background:

• Intensive longitudinal data (ILD) are crucial for studying dynamic relationships between variables.
• Timescale mismatch between variables in ILD presents a significant analytical challenge.
• Existing dynamic structural equation modeling (DSEM) approaches like partial-path and average-score models have limitations.

Purpose of the Study:

• To evaluate existing DSEM models for timescale mismatched variables.
• To assess the performance of improved DSEM approaches: full-path, factor, and adjusted factor models.
• To provide methodological guidance for analyzing ILD with timescale mismatches.

Main Methods:

• Simulation studies (Study 1, 2-1, 2-2) comparing model performance under various conditions.
• Evaluation of partial-path, average-score, full-path, factor, and adjusted factor models.
• Application of models to empirical data with timescale mismatched variables (Study 3).

Main Results:

• The full-path model superiorly captured dynamic interactions and time-specific effects compared to the partial-path model.
• The factor model provided accurate estimates for timescale mismatched variables, unlike the biased average-score model.
• The adjusted factor model offered marginal improvements over the factor model when regression effects were substantial.

Conclusions:

• Timescale mismatch is a critical issue in ILD analysis.
• The full-path and factor models are recommended for analyzing dynamic relations with timescale mismatches.
• This research offers valuable insights for data collection and analysis strategies in ILD studies.