Weighted Batch Means for Confidence Intervals in Steady-State Simulations

摘要: We propose a new procedure for providing confidence-interval estimators of the mean covariance-stationary process. The procedure, modification method batch means, is an improvement over existing methods when process displays strong correlation and comparatively small number observations available. We assign weights to within batch. are determined by order time-series model fit its estimated parameters. For given parameters, minimize variance weighted point estimator mean; formed these optimal applied unbiased. identification estimation parameters bring in bias. Formulas ARp MA1 processes. Experiments were conducted on AR1, AR2, MA2 processes as well M/M/1 queue delay periodic inventory system cost Results show that with many lags, typical queue-delay processes, achieved coverage constructed interval closer nominal levels than unweighted-batch-means method. average half-length not greatly affected batches, nonnormality means severe.