A1011
Title: Functional-coefficient models for multivariate time series in designed experiments: Applications to brain signals
Authors: Paolo Victor Redondo - King Abdullah University of Science and Technology (Saudi Arabia) [presenting]
Raphael Huser - King Abdullah University of Science and Technology (Saudi Arabia)
Hernando Ombao - King Abdullah University of Science and Technology (KAUST) (Saudi Arabia)
Abstract: To study the neurophysiological dynamics of attention deficit hyperactivity disorder (ADHD), clinicians use multichannel electroencephalography (EEG) which records neuronal electrical activity in the cortex. The most commonly-used metric in ADHD is the theta-to-beta power ratio (TBR) which is derived from the spectrum of the EEGs. However, initial findings for this measure have not yet been replicated in other studies. Instead of focusing on spectral power, this paper develops a novel model for investigating dependence between channels in the entire network. Although dependence measures such as coherence and partial directed coherence (PDC) are well explored in studying brain connectivity, these measures only capture linear dependence. Moreover, in designed clinical experiments, it is observed that these dependence measures can vary across subjects even within a homogeneous group. Hence, to address these limitations, we propose the mixed-effects functional-coefficient autoregressive (MX-FAR) model. The advantages of MX-FAR are the following: (1.) it captures non-linear dependence between channels; (2.) it is nonparametric and hence flexible; (3.) it can capture differences between groups; (4.) it accounts for variation across subjects; (5.) the framework easily incorporates well-known inference methods from mixed-effects models. Finally, we showcase the MX-FAR model through numerical experiments and report novel findings from the analysis of EEG signals in ADHD.
