An Efficient Method of Estimating Confidence Intervals for Use in Simulation-Optimization

In many applications of simulation-optimization, when comparing two or more alternatives, it is crucial to be able to estimate the confidence intervals on the outputs of interest with a reasonable level of accuracy. This acuracy has often been tested by the closeness of the coverage of the estimated confidence interval to the intended coverage. In this paper two variations to the Batch-Means Method of estimating the confidence intervals are presented and their performance are compared with the original method. The results indicate that the Batch Means Method modified by factors obtained by a second order autoregressive method is superior to the original and the one modified based on factors obtained from autocorrelation analysis.