Abstract
Computer simulation has been proved to be a very powerful tool in evaluating complex simulation.A fundamental problem of simulation output analysis is determining the quality of performance-measure estimators. A prototype problem is estimating the variance of the sample mean from a covariance-stationary process sample Y1,Y2,…,Yn. Most variance estimators of the mean of Y,including batch mean estimators,assume the memory space is big enough to store all data. This thesis pursues the optimal variance variance estimator by assuming the memory space is finite and the sample size is not known in advance. Through the verification of methodology implementation, our variance estimator is shown to perform better than the existing estimator.