Estimating Time Averages via Randomly-Spaced Observations.
Technical summary rept.,
WISCONSIN UNIV-MADISON MATHEMATICS RESEARCH CENTER
Pagination or Media Count:
In many stochastic systems, one is interested in estimating steady-state expected values. When Monte Carlo simulation is used to estimate such parameters, an assessment of accuracy, in the form of confidence intervals, is often required. Most procedures for producing such confidence intervals require that the simulation be sampled so that the time increments between observations are all equal. This is difficult to accomplish in a discrete-event simulation, since the clock which drives the simulation is incremented in a random fashion. To estimate continuous-time averages via randomly-spaced observations of discrete-event systems, the authors develop a point-process framework and use it to generalize both regenerative and stationary-process oriented simulation methodologies. They give consistent estimators, central limit theorems, and an effective bias-reducing jackknife. The impact of indirect estimation of transaction customer averages is discussed.
- Statistics and Probability