摘要
幂变换是多元线性回归分析中数据预处理的有效办法之一.以胰岛素注射治疗糖尿病为例,探讨幂变换实用的条件、研究方法及研究结论,并将幂变换处理后的变量,用于线性回归分析,研究医学问题.具体就是通过对变量数据描述性分析了解数据的特点,相关分析及聚类分析确定胰岛素初始剂量的重要影响因素,幂变换对数据进行预处理,继而进行回归分析,并对比变换前初步回归分析的结果,确定最终以尿蛋白分类标准,RI用量与血糖的对数之间的线性回归模型.
Power transform is one of the data pre-treatments in the linear regression analysis. The article cites the example on diabetes mellitus patients with insulin cure, in order to make an inquiry into studying condition,method and result by the way of power transform and use the linear regression analysis to study medicine question. The paper describes the special feature on data,decides on the key influence factors by correlation analysis and cluster analysis, handles data in power transform method in advance,and does regression analysis so as to determine the linear regression analysis modle between Insulin dosage and bloood sugar logarithm on the basis of urinary protein classification against previously.
出处
《数学的实践与认识》
CSCD
北大核心
2010年第10期131-140,共10页
Mathematics in Practice and Theory
基金
河南省教育厅自然科学研究计划项目(2008C110006)
关键词
相关分析
正态性检验
聚类分析
幂变换
回归分析
correlation analysis
test of normality
cluster analysis
power transform
multiple regression analysis
