建立以游程论为基础的控制图判断准则

控制图是统计质量管理的基本工具。运用控制图的重要问题在于弄清工艺波动是否归因于随机干扰。但这只能间接地从过程的非随机状态即控制图上点子的非随机排列来加以判断。通常基于多项分布建立起来的判断准则,由于忽视了点子出现顺序的信息,检出异常原因的能力很低。Schewhavt,Mostel-ler,Swed 和 Eisenhart 对此曾有所改进,但仍然丢掉了许多信息。作者在本文中试图阐明运用游程论判断控制图的优越性,并基于不同种类不同长度游程数的概率函数建立一套非随机状态的判断准则。对于同样的显著水平,按照这套准则,一般可以用比传统方法较小的样本及早揭示可能的异常原因。

The Establishment of the Criteria of Judgement for the Control Chart Based on the Theory of Runs

A control chart is a basic tool for statistical quality control.In using the control chart,it is important to know whether a process is operating in random variations.However,it can be demonstrated only indirectly by indicating when a process,that appears equiva- lently in pattern of points on control chart,is not random.Generally, as for the criteria judging non-random manner based on multinomial distribution,information of the order of the points presented on a control chart is usually entirely neglected.Such disregard results in low power of detection of assignable causes,Schewhart,Mosteller, Swedand Eisenhart made the judgement better,but yet much available information were lost,In this paper,the author tries to demonstrate the advantage of using the the theory of runs for judging control chart,and intends to base a package of criteria for jndging non-ran- dora manner that it is possible,on the same level of significance,to discover the occurrence of assignable causes by taking smaller sample than that of the traditional methods,resulting in practical and eco- nomical significance.