摘要
传统计算非正态分布过程能力指数最经典的方法--Clement方法,其最大的缺点是必须有足够多的观测样本才能得到较为准确的结果。文章利用加权标准差可将非正态过程分解成两个正态过程的思想,结合样本估计相关理论构建了一种基于加权标准差的过程能力指数。新指数无论是在小样本还是大样本的情况下,都比Clement方法估计结果的准确性更高,且在此方法基础上构建的Bootstrap置信区间的真实值覆盖率均远远高于同等条件下Clement方法构建的置信区间。
Clement is the most classic method to calculate non-normally distributed process capability index,but this method has the biggest disadvantage that there must be enough observed samples to get accurate results.This paper makes use of the idea that non-normal processes can be decomposed into two normal processes by using weighted standard difference,and also combines with the related theory of sample estimation to construct a process capability index based on weighted standard deviation.The new index is more accurate than the Clement method in the case of both small and large samples,and the true value coverage rate of the Bootstrap confidence interval constructed based on the proposed method is much higher than that constructed by the Clement method under the same conditions.
作者
颜斌
王斌会
徐锋
Yan Bin;Wang Binhui;Xu Feng(Economic Management Experimental Teaching Center,Jinan University,Guangzhou 510632,China)
出处
《统计与决策》
CSSCI
北大核心
2020年第7期10-16,共7页
Statistics & Decision
基金
国家社会科学基金一般项目(16BTJ035)。
