Abstract
We introduce the binary bootstrap for inference with autoconelated binary data. Weempirically evaluate the standard eirors and confidence intervals created with the binary bootstrap using four stochastic process with known results: Bernoulli trials, first-order Markov processes, and long customer delays in M/M/l and D/M/10 queues. The binary bootstrap has certain advantagesover the conventional batch means method for creating confidence intervals on the probability oflong delay using only one run of a discrete-event simulation