|
Abstract:
|
Sequences of action potentials from neurons, known as spike trains, can be modeled as point processes. The assumption that spike times follow a Poisson process is insufficient when modeling within-trial firing behavior. Dependency upon past firing events, trial-dependent effects, and correlation with other variables can lead to significant deviations from the simple Poisson process. In order to incorporate the past spiking behavior of the neuron, the inhomogeneous Markov interval process, has previously been proposed. This model is expanded to accommodate other effects, such as latency and excitability effects, correlation with recorded oscillatory behavior, and reaction to activity in other neurons. A Gibbs sampling procedure is used to estimate the effect of each aspect, applying the mechanism of Bayesian Adaptive Regression Splines (BARS) to obtain smoothed estimates of effects.
|
