Determining simulation sample size for given precision requirements

摘要

Simulation credibility evaluation concerns rncomparisons of the performance measures between rnsimulation system and its rnreference counterpart, In fact, the required rnprecision is known as prior information, which brings rnthe problem of determining rnsimulation run length according to the estimation of rnperformance measures and the required precision. A rnfeedback method was rnproposed from model analysis to data collection. For rntransient performance measures, operating rncharacteristic curves were employed rnto calculate the simulation replications to detect rnthe difference of means in parameter range. For rnsteady-state performance measures,rntwo phase method was put forward to satisfy the warm-rnup sizes at the first phase, and a run length for the rnrelative- or rnabsolute-precision requirement at the second phase. rnSome empirical experiments illustrate the validity of rnthe proposed method.