Control variate models for sensitivity estimates of repairable item systems

  • Authors:
  • Mohamed A. Ahmed;Donald Gross;Douglas R. Miller

  • Affiliations:
  • Department of Operations Research, School of Engineering and Applied Science, George Washington University, Washington, DC;Department of Operations Research, School of Engineering and Applied Science, George Washington University, Washington, DC;Department of Operations Research, School of Engineering and Applied Science, George Washington University, Washington, DC

  • Venue:
  • WSC '87 Proceedings of the 19th conference on Winter simulation
  • Year:
  • 1987

Quantified Score

Hi-index 0.00

Visualization

Abstract

We previously developed a new modeling idea for comparing infinite-source, ample-server models (∞/∞) and finite-source, finite-server models (f/f). The comparison provides an efficient estimate of the error when approximating an f/f system with an ∞/∞ system and allows the analytical solution of the ∞/∞ model to be used as a control variate. We show that using ∞/∞ models as control variates for f/f systems can be an effective variance reduction technique for sensitivity estimates (gradients, Hessians). We also compare this to a method using concomitant control variates for the sensitivity estimates. Using these sensitivity estimates we will be able to determine more efficiently when ∞/∞ models are good approximations for f/f systems.