A Granger Causality Measure for Point Process Models of Neural Spiking Activity

A network-centric view of brain function that is becoming more widely accepted would benefit from the directional interaction information that occurs between multiple neurons. Granger causality has been used previously to address this need, but previous methods can only operate on continuos-value data sets. This prevents it from being directly applied directly to neural spike train data so previous attempts have involved smoothing, binning, etc. to address this. Recently a point process framework that enables Granger causality to be applied to point process data was proposed. The newly proposed framework was used to investigate network interactions proposed in previous assignments to verify that the correct minimal generative model graph could be recovered from just the raw simulated neural spike train data. The interactions and non interactions present in the simulated networks were indeed recovered.