Inference for the diffusion models of neuronal activity (Q788449): Difference between revisions

Diffusion models of neuronal activity are analyzed from the view point of obtaining appropriate estimators. The two basic mechanisms that are widely used are: (i) the perfect integrator in which the infinitesimal drift and variance are essentially characterized by two constants \(\mu\) and \(\sigma\) ; (ii) the leaky integrator in which the infinitesimal drift and variance are given by \(\mu(x)=-x/\tau +\mu\), \(\sigma(x)=\sigma>0\) where \(\tau>0.\) The perfect integrator has the advantage that it leads to the Ornstein- Uhlenbeck process. The statistical analyses of these models have all along been based on the assumption that samples of interspike intervals are available for analysis. In the paper under review, it is assumed that sample values of the membrane potential for some instant between consecutive spikes are available. Explicit expressions are provided for the estimators \({\hat \mu}\) and \({\hat \sigma}\) in the case of perfect integrator, and \({\hat \mu}\), \({\hat \sigma}\) and \({\hat \tau}\) in the case of leaky integrator.