Estimating both directed and undirected contemporaneous relations in time series data using hybrid-group iterative...

Area of Science:

• Behavioral Sciences
• Psychology
• Social Sciences

Background:

• Intensive longitudinal data (ILD) analysis is crucial for understanding individual processes over time.
• Existing methods for modeling contemporaneous relations in vector autoregression (VAR) models present a forced choice between undirected or directed relations, potentially introducing bias.
• This limitation necessitates a more flexible approach to accurately capture instantaneous relationships within ILD.

Purpose of the Study:

• To introduce hybrid-group iterative multiple model estimation (GIMME), a novel data-driven method for analyzing intensive longitudinal data (ILD).
• To provide a unified modeling framework that accommodates both directed and undirected contemporaneous relations alongside standard VAR lagged relations.
• To offer a solution that mitigates bias and improves inference in the analysis of complex temporal dynamics.

Main Methods:

• Development and application of hybrid-group iterative multiple model estimation (GIMME), a data-driven approach.
• Integration of vector autoregression (VAR) for lagged relations with flexible modeling of contemporaneous relations (directed and undirected).
• Validation using both simulated and empirical intensive longitudinal datasets.

Main Results:

• Hybrid-GIMME demonstrates robustness in recovering contemporaneous relations from intensive longitudinal data.
• The method effectively integrates different types of contemporaneous relationships within a single analytical framework.
• Performance is particularly strong when a sufficient number of time points per individual is available.

Conclusions:

• Hybrid-GIMME offers a significant advancement in the analysis of intensive longitudinal data (ILD) by providing a unified approach to modeling contemporaneous relations.
• This novel method enhances the accuracy of uncovering temporal dynamics and reduces potential biases inherent in previous approaches.
• The findings support the utility of hybrid-GIMME for exploratory analysis of complex relationships in ILD, especially with rich temporal data.