Research
Seminars
Department Series
[an error occurred while processing this directive]
The Variance of Time Averages in Markovian Event Simulation |
|
Dr. Winfried Grassmann Department of Computer Science University of Saskatchewan |
|
In simulation, one typically uses time averages to estimate performance measures, such as expected queue lengths or blocking probabilities. To find how accurate these estimates are, one needs the variances of the estimators. We show how to obtain these variances analytically in the context of Markovian event systems. A discrete event system is called a Markovian event system if all rates depend only on the present state, and not on the past. Analytical solutions let us gain insight into the factors affecting the variance, and, in this way to estimate the simulation run length needed to obtain estimates of prescribed accuracy.
Professor Grassmann worked at the then Swiss flag carrier Swissair from 1968-1968. He was awarded his Ph.D. in 1968, and joined the Department of Computer Science in January 1969. He has written 3 books, one with Professor Tremblay (please do not hold this against him) and numerous articles, mostly on queueing theory and simulation.
[an error occurred while processing this directive]