Mean Value Estimation from Digital Computer Simulation

Publisher Website
Google Scholar
Abstract
Terminating and nonterminating stochastic processes are studied. If digital simulation is used, either a time-slicing or an event-sequencing technique may be used to obtain an estimate of some measure (mean value in our case) of system performance. We consider the variance of the estimate to be the basic measure for determining the goodness of the estimate. Three points are examined: (1) unbiasedness of the estimate, (2) efficiency of the estimate, and (3) variance reduction techniques. For the nonterminating case a large class of steady-state processes having a negative exponential covariance function is studied and curves and specific examples are derived. Replication is shown to be generally superior either to increasing the length of a run or to reducing the sampling interval. For the terminating case, a general expression for efficiency is derived and replication is shown to be superior to decreasing the sampling interval. All results are equally valid for discrete or continuous processes and do not consider the cost trade-offs.

This publication has 0 references indexed in Scilit: