Analysis and modeling of quasi-stationary multivariate time series and their application to middle latency auditory evoked potentials

A methodological framework for analyzing and modeling of multivariate data is introduced. In a first step, a cluster method extracts data segments of quasi-stationary states. A novel cluster criterion for segment borders is introduced, which is independent of the number of clusters. Its assessment reveals additional robustness towards initial conditions. A subsequent dynamical systems based modeling (DSBM) approach focuses on data segments and fits low-dimensional dynamical systems for each segment. Applications to middle latent auditory evoked potentials yield data segments, which are equivalent to well-known waves from electroencephalography studies. Focussing to wave Pa, two-dimensional dynamical systems with common topological properties are extracted. These findings reveal the common underlying dynamics of Pa and indicate self-organized brain activity.