This paper discusses the efficiency of various batching methods for estimating performance parameters from steady-state simulation output, e.g., the steadystate mean. Our primary focus is on issues related to computational and storage requirements of batching methods such as batch means, overlapping batch means, and standardized time series. We also examine the important question of determining the proper batch size for a given estimation problem — how much effort do batch-size algorithms require?
David Goldsman, Bruce W. Schmeiser