Abstract
When simulating a stochastic system, simulationists often are interested in estimating various steady-state performance measures. The classical point estimator for such a measure involves simply taking the time average of an appropriate function of the process being simulated. Since the simulation can not be initiated with the (unknown) steady-state distribution, the classical point estimator is generally biased. In the context of regenerative steady-state simulation, a variety of other point estimators have been developed in an attempt to minimize the bias. In this paper, we provide an empirical comparison of these estimators in the context of four different continuous-time Markov chain models. The bias of the point estimators and the coverage probabilities of the associated confidence intervals are reported for the four models. Conclusions are drawn from this experimental work as to which methods are most effective in reducing bias.
- Cash, C. R., Dippold, D., Long, J., Nelson, B., and Pollard, W. 1992. Evaluation of tests for initial conditions bias. In Proceedings of the 1992 Winter Simulation Conference. IEEE Computer Society, Press, Los Alamitos, Calif., 577--585. Google Scholar
- Glynn, P. W. 1984. Some asymptotic formulas for markov chain with applications to simulation. J. Stat. Comput. Simul. 19, 97--112.Google Scholar
- Glynn, P. W. 1989. A GSMP formalism for discrete event systems. Proc. IEEE 77. 14--23.Google Scholar
- Glynn, P. W. 1994. Some topics in regenerative steady-state simulation. Acta Appl. Math. 34, 225--236.Google Scholar
- Glynn, P. W. 1995. Some new results on the initial transient problem. In Proceedings of the 1995 Winter Simulation Conference. IEEE Computer Society Press, Los Alamitos, Calif. (Arlington, Va.). 165--170. Google Scholar
- Glynn, P. W. and Heidelberger, P. 1990. Bias properties of budget constrained Monte Carlo simulations. Oper. Res. 38, 801--814. Google Scholar
- Glynn, P. W. and Heidelberger, P. 1991a. Analysis of initial transient deletion for replicated steady-state simulations. Oper. Res. Lett. 10, 437--443.Google Scholar
- Glynn, P. W. and Heidelberger, P. 1991b. Analysis of parallel, replicated simulations under a completion time constraint. ACM Trans. Model. Comput. Simul. 1, 3--23. Google Scholar
- Glynn, P. W. and Heidelberger, P. 1992a. Analysis of initial transient deletion for parallel steady-state simulations. SIAM J. Sci. Stat. Comput. 13, 909--922. Google Scholar
- Glynn, P. W. and Heidelberger, P. 1992b. Jackknifing under a budget constraint. ORSA J. Comput. 4, 226--234.Google Scholar
- Goldsman, D., Schruben, L., and Swain., J. 1994. Tests for transient means in simulated time series. Naval Res. Logist. Quart. 41, 171--187.Google Scholar
- Henderson, S. G. and Glynn, P. W. 2001. Regenerative steady-state simulation of discrete-event systems. ACM Trans. Model. Comput. Simul. 11, 313--345. Google Scholar
- Iglehart, D. L. 1975. Simulating stable stochastic systems, V: Comparison of ratio estimators. Naval Res. Logist. Quart. 22, 553--565.Google Scholar
- Meketon, M. and Heidelberger, P. 1982. A renewal theoretic approach to bias reduction in regenerative simulations. Manage. Sci. 26, 173--181.Google Scholar
- Nelson, B. 1992. Initial-condition bias. In Handbook of Industrial Engineering, 2 ed., G. Salvendy, Ed. Wiley, New York.Google Scholar
- Schruben, L. 1982. Detecting initialization bias in simulation output. Oper. Res. 30, 3, 151--153.Google Scholar
- Schruben, L., Singh, H., and Tierney, L. 1983. Optimal tests for initialization bias in simulation output. Oper. Res. 31, 6, 1167--1178.Google Scholar
- White, K. P., J. 1997. An effective truncation heuristic for bias reduction in simulation output. Simulation 69, 6, 323--334. Google Scholar
- White, K. P., J., Cobb, M. J., and Spratt, S. C. 2000. A comparison of five steady-state truncation heuristics for simulation. In Proceedings of the 2000 Winter Simulation Conference. IEEE Computer Society Press, Loss Alamitos, Calif., 755--760. Google Scholar
Index Terms
- Empirical performance of bias-reducing estimators for regenerative steady-state simulations
Recommendations
A Coupling-Based Analysis of a Multiclass Retrial System with State-Dependent Retrial Rates
Queueing Theory and Network ApplicationsOn the theoretical comparison of low-bias steady-state estimators
The time-average estimator is typically biased in the context of steady-state simulation, and its bias is of order 1/t, where t represents simulated time. Several “low-bias” estimators have been developed that have a lower order bias, and, to first-...
How Hard are Steady-State Queueing Simulations?
Special Issue on Don IglehartSome queueing systems require tremendously long simulation runlengths to obtain accurate estimators of certain steady-state performance measures when the servers are heavily utilized. However, this is not uniformly the case. We analyze a number of ...
Comments