Abstract
The Kolmogorov-Smirnov test is used to determine whether the output distributions of two simulated systems are significantly different. The advantages of using this statistical procedure, and the danger in making comparisons between the systems with only the mean as a critical performance measure, are dem onstrated.
Two queuing systems were simulated and results analyzed first by comparing means and, then, distributions. Results showed that, even if the means of two alternative systems are equivalent, the systems may still prove to be significantly different from each other if the entire distribution is examined.
Keywords
Get full access to this article
View all access options for this article.
References
1.
Amer, P.
"Selection Procedures Based on Proportion Parameters ." Winter Simulation Conference Proceedings (1980 ), 3-8.
2.CACI SIMSCRIPT II 5 Programming Language . CACI, Inc.
Los Angeles, Calif. (1983 ).
3.
Gibbons, J.D.
Nonparanietric Methods for Quantitative Analysis . Holt , New York, N.Y. (1976 ).
4.
Gross, D.
and
Harris, C.M.
Fundamentals of Queuing Theory . Wiley , New York, N.Y. (1974 ).
5.
Iglehart, D.L.
"Simulating Stable Stochastic Systems, VI: Quantile Estimation ." Journal of the ACM
23 (1976 ), 347 -360 .
6.
Jucker, J.V.
and
Gomez, J.G.
"Policy-Comparing Simulation Expenments: Design and Analysis "Decision Sciences
6 (1975 ), 631 -645 .
7.
Kelton, W.D.
and
Law, A.M.
" An Analytical Evaluation of Alternative Strategies in Steady-State Simulation ." Operations Research
32 (1984 ), 169 -184 .
8.
Mamrack, S.A.
and
Amer, P.D.
"Estimating Confidence Intervals for Simulations of Computer Systems ." SIMULATION
35 (1980 ), 199 -205 .
9.
Seila, A.F.
"A Batching Approach to Quantile Estimation in Regenerative Simulations ." Management Science
28 (1982 ), 573 -581 .
10.
Siegel, S.
Nonparametric Statistics for the Behavioral Sciences . McGraw-Hill , New York, N.Y. (1956 ).