| Abstract: | An algorithm is presented which combines the techniques of statistical simulation and numerical integration, thus furnishing improved estimates of cumulative distribution functions. The method uses statistical estimation techniques to form a statistic possessing an approximate normal distribution. A post-stratification sample is used to form a control variable correction for the original stratified numerical estimate, and this combination results in a competitor for stratified Monte Carlo sampling. Examples are presented for known statistical distributions and for an actual electronics system. A branch and bound algorithm, which can reduce the amount of computations necessary for both control variable stratified Monte Carlo sampling, is developed for monotone functions. |
|---|