Date of Award

12-1992

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Department of Operational Sciences

First Advisor

Kenneth W. Bauer, Jr., PhD

Abstract

Data truncation is a commonly accepted method of dealing with initialization bias in discrete-event simulation. Algorithms for determining the appropriate initial-data truncation point for univariate and multivariate output are proposed. The techniques entail averaging across replications and estimating a steady-state output model in a state-space framework. Using the estimated model, Multiple Model Adaptive Estimation (MMAE), which uses Kalman filters with different parameter vectors, is applied. Based on the filters' residuals, the conditional probabilities of each filter's specific parameter vector being correct are determined. The MMAE parameter estimates are the probabilistic- weighted average of the filters' assumed parameter vectors. The estimated truncation point is selected when a vector norm of the difference between an estimate of the steady-state mean vector and the MMAE time-varying estimated mean vector are within a small tolerance. A Monte Carlo analysis using data generated from discrete-event simulations are used to evaluate the techniques. The evaluation criteria include the ability to accurately estimate and to construct reliable confidence regions for the vector of the response means based on the truncated sequences.

AFIT Designator

AFIT-DS-ENS-92-1

DTIC Accession Number

ADA256611

Comments

The author's Vita page is omitted.

Download

Share

COinS