Date of Award

3-21-2013

Document Type

Thesis

Degree Name

Master of Science in Operations Research

Department

Department of Operational Sciences

First Advisor

Mark A. Friend, PhD.

Abstract

At the heart of most modeling issues is a focus on variance reduction. Experimental designs are chosen based on both efficiency and a variety of variance based criteria. In many situations due to cost, time and availability issues it is beneficial to produce metamodels of simulations. Experimental designs for the region of operability are constructed to collect the simulation output required to construct representative metamodels. Independently, the method of control variates is a well established technique often employed to reduce variance in discrete event simulations. This thesis explores the variance reduction benefits that can be obtained by combining optimal experimental designs with control variates in multipopulation simulation experiments when constructing simulation metamodels. A variety of variance measures of effectiveness are used to demonstrate the theoretical benefits obtained by this approach. In addition, a randomly selected data set from within the design region is used to demonstrate the practical application and reduction of predictive variance obtained using this methodology.

AFIT Designator

AFIT-ENS-13-M-11

DTIC Accession Number

ADA584605

Download

Share

COinS