“…The statistical modelling of neural spike trains falls under the general framework of point processes. Among point processes, the (inhomogeneous) Poisson process is a popular choice [1,14], but its theoretical properties fail to address biological phenomena such as refractoriness and burst spiking activities [15]. Furthermore, the Poisson process has very limited multivariate generalizations in terms of dependence structure [12,9,4,11,6,7].…”