|
We investigated serial event history in time series of action potentials as generated by spiking neurons. In periods of stationary firing we found significant short-ranged serial correlation of the inter-event intervals in cortical pyramidal neurons of the rat and in output neurons from the mushroom body of the honeybee. This result is consistent with earlier findings in peripheral sensory systems, e.g. in the weakly electric fish, and is believed to be a common feature of neurons that exhibit spike frequency adaptation.
To incorporate serial interval statistics into the framework of a point process model we define a class of autoregressive models of inter-event intervals. We fit the marginal interval distribution to experimental distributions and use maximum likelihood estimators to obtain the regressive model parameters from neuronal data. These models help us to formulate improved predictions, e.g. on count statistics and cross-correlation statistics, which are required for the statistical analysis of multi-dimensional spiking data. |
|