Chapter 6 - State Space Modeling of Neural Spike Train and Behavioral Data

Abstract

State space modeling is an established framework for analyzing stochastic and deterministic dynamical systems that are measured or observed through a stochastic process. This highly flexible paradigm has been successfully applied in engineering, statistics, computer science, and economics to solve a broad range of dynamical systems problems. The revolution in neuroscience recording technologies in the last 20 years has provided many novel ways to study the dynamic activity of the brain and central nervous system. These technologies include multielectrode recording arrays functional magnetic imaging electroencephalography and magneto encephalography, diffuse optical tomography, calcium imaging, and behavioral data. Because a fundamental feature of many neuroscience data analysis problems is that, the underlying neural system is dynamic and is observed indirectly through measurements from one or a combination of these different recording modalities, the state space paradigm provides an ideal framework for developing statistical tools to analyze neural data. Neural spiking activity recorded from single or multiple electrodes is one of the principal types of data recorded in neurophysiological experiments. Because neural spike trains are time series of action potentials, point process theory has been shown to provide an accurate framework for modeling the stochastic structure of single and multiple neural spike trains.