Principal component analysis based on a subset of variables: variable selection and sensitivity analysis
Principal component analysis based on a subset of variables: variable selection and sensitivity analysis
复制标题
基于变量子集的主成分分析：变量选择和敏感性分析
DOI：
10.1080/01966324.1997.10737430
复制
发表时间：
1997
影响因子：
--
通讯作者：
Y. Mori
中科院分区：
文献类型：
--
作者：
Y. Tanaka;Y. Mori
基于 Rao（1964）工具变量主成分分析和 Robert 和 Escoufier（1976）使用 RV 系数的方法的思想，提出了改进的主成分分析来导出主成分，这些主成分被计算为变量子集的线性组合，但可以很好地再现所有变量。应用向后消除程序来找到合适的变量子集序列，并开发了敏感性分析方法来检测改进的主成分分析中的有影响的观测值和有影响的变量。显示数值示例来说明所提出的程序的性能。
SYNOPTIC ABSTRACTA modified principal component analysis is proposed to derive principal components which are computed as linear combinations of a subset of variables but which can reproduce all the variables very well, based on the ideas of Rao(1964)'s principal component analysis of instrumental variables and Robert and Escoufier(1976)'s approach using the RV-coefficient. A backward elimination procedure is applied to find a suitable sequence of subsets of variables, and methods of sensitivity analysis are developed for detecting influential observations and influential variables in the modified principal component analysis. Numerical examples are shown to illustrate the performance of the proposed procedure.