On Confidence Intervals for Cyclic Regenerative Processes.
Abstract
Simulation is a commonly used method of analysis for studying complex stochastic systems. Often, the parameter of interest to the simulator can be estimated by more than one quantity. When more than one estimator exists, it is desirable to use the more stable estimate, namely the one with the lesser variance. In this paper, the authors consider a class of stochastic processes which enjoy cyclic regenerative structure - such systems often arise, for example, in analysis of queues. They study a family of estimators and determine precise conditions under which the estimators are asymptotically valid. They also obtain a closed-form solution for the minimum variance estimate in the family, and prove that this estimator will often be superior to the standard regenerative estimator for the simulation.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1983
- Accession Number
- ADA127761
Entities
People
- Peter W. Glynn
Organizations
- University of Wisconsin–Madison