A harmonic analysis approach to simulation sensitivity analysis

Simulation sensitivity analysis is an issue faced by simulation practitioners analyzing complex systems. This paper presents steady-state simulation expected response gradient and Hessian estimators, using harmonic analysis. These estimators exploit the orthogonality of the harmonic basis and require just one simulation run, independent of the number of input parameters. Convergence properties and implementation issues for the estimators are discussed. Examples are included to illustrate the performance of the estimators.