Parallel Sequential Estimation of Quantiles During Steady State Simulation

Abstract

Simulation results are often limited to mean values, even though this provides very limited infor- mation about the analyzed systems' performance. Quantile analysis provides much deeper insights into the performance of simulation system of interest. A set of quantiles can be used to approximate a cumulative distribution function, providing full information about a given performance characteristic of the simulated system. In this paper, we will present two methods for parallel sequential estimation of steady state quantiles. The quantiles are estimated using simulation output data from concur- rently executed independent replications. They are calculated sequentially and on-line, i.e. during simulation, to ensure that the results are produced to a speci ed accuracy. The set of quantiles to be estimated can be automatically determined using e cient estimation as a criterion.