AbstractThis paper discusses The problem of generating confidence limils from the output of a lingle run on steady state characteristics of simulation models. The only assumption made aboul the model is that steady state behavior exists. The method is based on spectral analysis techniques. The variance estimate neceasary for the formation of the confidence interval is obtained by eatimating The spectral density at zero frequency. This estimation is accomplisbed through the application of regression techniques 10 the logarithm of the averaged periodogram in the low frequency region. The method is applied 10 batched data in order 10 keep storage requirements low. The implementation of this method as part of a practical run length control procedure is also discussed.