摘要
使用均匀设计试验时有两点须要注意的问题 :1 .均匀设计表中有若干个试验方案有列共线现象 ,对其试验结果进行回归分析时 ,正则方程组是病态的 ,将得到不易察觉的错误结论 ;2 .与回归型设计或正交表设计不同 ,一组均匀设计的试验结果有多个回归方程可供选择 ,其回归系数不一定反映真实的因子效应 .因此 ,在使用均匀设计研究有重要经济背景或有关人身安全的课题时应极端谨慎 .本文将避开数学论证 ,以直观和可以重复验证的事例探讨有列共线现象的试验方案 。
There are two topics that will be noticed when utilizing the Uniform Design: 1. In the tables of Uniform Design, there are several experiment planes that have column linear equivalent phenomena. When proceeding regression analysis on the result of the experiments, the normal equation group is ill conditioned and will gain an imperceptible false conclusion. 2. Different with the regression designs and orthogonal type designs, in a series of uniform design experiments, there are several kinds of regression equations that can be selected. It is unlikely that the regression coefficiency reflects the factor effect in truth. So in the research of topics which has important economic background or concerns with the security of people, one should be extremely careful. This paper avoided the mathematical argue discussed the above topics with the visualized, repeatable examples,listed out the experiment planes which present column linear equivalent phenomena. The author established the optimal factor subset method for searching the professional acceptable regression equation.
出处
《淄博学院学报（自然科学与工程版）》
2002年第1期57-61,共5页
Journal of Zibo University(Natural Sciences and Engineering)
关键词
均匀设计试验
列共线
因子效应
最佳因子集
回归方程
uniform design
column linear equivalent
regression method of the optimal factor subset
