Generalized sensitivity analysis of ergodic stochastic systems (Q1117615)

The tools for sensitivity analysis of stochastic systems whose behavior is described by ergodic stationary processes are presented. Three kinds of generalized estimates for performance sensitivities of these systems are introduced. These estimates, based on a use of time averages, are given in general operator form. Their convergence conditions and rate of convergence are obtained. It is shown that perturbation analysis estimates are generally preferable to crude Monte Carlo estimates.