均匀设计实验资料最优回归子集的构建

Creating Best Regression Subsets for the Experimental Data of the Uniform Design

目的本文针对均匀设计实验资料的回归分析结果不唯一的事实,提供构建均匀设计实验资料最优回归子集的策略。方法 拟合包含不同数目自变量的回归方程,运用SAS语言编写的程序对方程进行假设检验以获得最大F值为判断标准寻找最优回归方程,并用一个真实的实验资料演示这一计算过程和结果。结果 能够对给定的均匀设计实验资料以较快的速度和较高的精确度获得"较优(甚至最优)"的回归方程。结论本文所提供的方法是构建均匀设计实验资料最优回归子集的一个有效方法。

Objective Based on the fact that the regression analysis results of the data collected from the Uniform Design are not unique,the author supplied a strategy of creating best regression subsets for the data. Methods A series of regression equations with different number independent variables are fitted, so that best regression subsets can be obtained according to the criteria of maximum F value by using SAS language. Furthermore, the calculating procedure and results are demonstrated by analyzing real experimental data. Results For the given data collected from the Uniform Design, a well -fitted (or best) regression equation can be obtained rapidly and accurately. Conclusion The method provided in this paper is an effective approch to creating best regression subsets for the data of the Uniform Design.