Proportion Estimation of Correlated Sequences

Abstract

Simulated estimates of proportions are needed in some simulation studies. We propose a quasi-in dependent method to determine the length of a simulation run so that the proportion estimates satisfy a pre-specified precision requirement. The method uses histogram approximation to generate confidence intervals with the desired precision. Because a histogram of the output sequence is constructed, several proportions can be estimated simultaneously without much extra effort. The assumptions required are that the stochastic pro cess output sequence is continuous and satisfies the ø-mixing conditions. An experimental perfor mance evaluation demonstrates the validity of the quasi-independent procedure with histogram ap proximation.