本篇論文提供一種「循序漸進的批量平均執」的技巧,用以求取在衡量多重辯量的穩定狀態模擬實驗中達到特定精確度的統計分析結果。這個技巧利用批量平均值的方法得到一連串近似獨立及近四來自相同方法分配的觀察值,然後運用一種循序漸進的程序來達到我們所希望的精確程度。於此,我們以涵蓋比率來作為判斷這種方法好壞的標準,同時選用單站等候系統(M/M/1)及雙站串聯系統(tandem queue)兩個模式來實際評估這個方法的優劣。在最後,我們列出了實驗研究的結果。
This paper proposed a sequential method based on batch means concept to achieve a specified accuracy in a multivariate steady-state simulation output analysis. It utillzes a multivariate batch means method to get some approximately i.i.d. observation vectors and then develops a sequential procedure to generate more observations to achieve a specified accuracy requirement. The coverage rate is used as the criterion to measure the performance of this method and a M/M/1 queue model and a tandem queue model are used to test it. Some empirical results are included.