Testing for group differences in multilevel vector autoregressive models

Multilevel Vector Autoregressive (VAR) models have become a popular tool for analyzing time series data from multiple subjects. Many studies aim to investigate differences in multilevel VAR models between groups, such as patients and healthy controls. However, there is currently no easily applicable method to make inferences about such group differences. Here, we present two standard tests for making such inferences: a parametric test and a nonparametric permutation test. We explain the rationale for both tests, provide an implementation based on the popular R-package mlVAR, and evaluate their performance in recovering group differences in scenarios resembling empirical research using a simulation study. Finally, we provide a fully reproducible R-tutorial on testing for group differences in a dataset of emotion measures using the new R-package mnet.