Derivative estimation via stochastic intensities: Event averages in queueing systems

We briefly describe sensitivity analysis methods for simulation via stochastic intensities and give applications on derivative estimation of event averages. We consider a process depending on a parameter and an event-counting process and obtain expressions for the derivatives of the expected number of events occurring while the process is in a given set. The necessary assumptions of these expressions to hold are also examined more specifically for queueing systems. The case of queues with renewal and nonhomogeneous Poisson arrivals is discussed in some detail.