The Determination of the Pilot Simulation Sample Size
初始模拟样本量的决定方法

With the use of simulation techniques, there exists a problem how to determinate the proper simulation sample size. The methods solving this problem include Absolute Error Method, Double Sampling Method, and so on. Among these methods, the pilot simulation runs have an important effect on the computed simulation sample size. Thus, it is important to choose the proper pilot simulation runs. If the pilot simulation sample is quite small, the computed sample size will vary very much. And on the contrary if the pilot simulation sample is quite large, time and recourse could be wasted. This paper addresses two important problems when choosing the proper pilot simulation runs: the variance of the pilot sample being in the steady state and the normality test of the sample mean being satisfied. It analyzes and recommends the tools and methods to judge and test these problems.