Testing for group differences in multilevel vector autoregressive models

Area of Science:

• Computational Statistics
• Time Series Analysis
• Multilevel Modeling

Background:

• Multilevel Vector Autoregressive (VAR) models are widely used for analyzing longitudinal data from multiple subjects.
• Investigating group differences (e.g., patients vs. controls) in these models is crucial but lacks standardized inferential methods.
• Existing methods for comparing multilevel VAR models across groups are not readily accessible or widely adopted.

Purpose of the Study:

• To introduce and evaluate novel statistical tests for inferring group differences in multilevel VAR models.
• To provide practical implementations of these tests within the R statistical environment.
• To assess the performance of the proposed methods in detecting group differences using simulation studies.

Main Methods:

• Development and explanation of a parametric test for group comparisons in multilevel VAR models.
• Development and explanation of a nonparametric permutation test for robust group comparisons.
• Implementation of both tests using the mlVAR R-package and a tutorial with the mnet R-package.

Main Results:

• The study successfully implemented and evaluated two distinct statistical tests for group differences in multilevel VAR models.
• Simulation studies demonstrated the performance of these tests in accurately recovering known group differences.
• The R-package mnet provides a reproducible framework for applying these methods to empirical emotion data.

Conclusions:

• The presented parametric and nonparametric tests offer accessible and reliable methods for group comparisons in multilevel VAR analysis.
• These tools enhance the ability to investigate inter-individual differences in dynamic network structures.
• The provided R implementations and tutorials facilitate the application of these advanced statistical techniques in psychological and neuroscientific research.