The authors examine a variant of the familiar batch means (BM) method for analysis of a stationary (simulation) process. The present spaced batch means (SBM) method attempts to reduce the bad effects of interbatch correlation by inserting spacers between the batches of observations. The authors present analytical examples in which SBM yields an estimator for the variance parameter that is less biased than the corresponding BM estimator. They also give analytical examples in which SBM improves coverage of confidence intervals for the mean of the process. Under negative serial correlation, SBM sometimes fares significantly better than BM; under positive serial correlation, SBM usually does only a bit better than BM.
