Analyzing event-related EEG data with multivariate autoregressive parameters

Methods of spatio-temporal analysis provide important tools for characterizing several dynamic aspects of brain oscillations that are reflected in the human scalp-detected electroencephalogram (EEG). The search to identify the dynamic connectivity of brain signals within different frequency bands, in order to uncover the transient cooperation between different brain sites, converges at the potential of multivariate autoregressive (MVAR) models and their derived parameters. In fact, MVAR parameters provide a whole battery of so-called coupling measures including classic coherence (COH), partial coherence (pCOH), imaginary part of coherence (iCOH), partial-directed coherence (PDC), directed transfer function (DTF), and full frequency directed transfer function (ffDTF). All of these approaches have been developed to quantify the degree of coupling between different EEG recording positions, with the specific aim to characterize the functional interaction between neural populations within the cortex. This work addresses the application of MVAR models to event-related brain processes, including different statistical approaches, and reviews most relevant findings in the expanding field of coupling analysis. Finally, we present several examples of coupling patterns associated with certain types of movement imagery.