Colloquium - Reza Ramezan (DSAS)

Speaker: Reza Ramezan (University of Waterloo)

Title: Bridging the Gap : Statistics, Neuroscience, and Spike Trains

Abstract: A scientific model is a conceptual/physical representation of a phenomenon while a statistical model is a probabilistic object which is usually derived based on the collected data from a phenomenon. While the adage “all models are wrong but some are useful’’ applies, it is still possible to combine the two approaches to improve model performance and model justification. This talk explains why it is important to do so, and introduces the Skellam Process with Resetting (SPR): a simple yet powerful and flexible model for neural spike trains which is biologically justified. Within a point-process framework, the SPR model is motivated by neural integration: a biological process through which nerve cells combine the information they receive from other nerve cells. It will be shown that certain parametrizations of the SPR result in popular models for inter-spike intervals (inter-arrival times) such as Gamma, Inverse Gaussian, etc. Most importantly, SPR has a flexible multivariate generalization for the analysis of simultaneously recorded spike trains and population coding. The model strengths and weaknesses will be highlighted through real data analysis and simulations.