基于稳健主成分回归的统计数据可靠性评估方法

统计研究 ›› 2011, Vol. 28 ›› Issue (8): 21-27.

基于稳健主成分回归的统计数据可靠性评估方法

卢二坡张焕明

出版日期:2011-08-15 发布日期:2011-08-08

An Evaluation Method of Reliability of Published Statistics Data Based on Robust Principal Component Regression Method

Lu Erpo & Zhang Huanming

Online:2011-08-15 Published:2011-08-08

摘要/Abstract

摘要： 稳健主成分回归（RPCR）是稳健主成分分析和稳健回归分析结合使用的一种方法，本文首次运用稳健的RPCR及异常值诊断方法，对2008年我国地区经济增长横截面数据可靠性做了评估。评估结果表明：稳健的RPCR方法能更好的克服异常值的影响，使估计结果更加可靠，并能有效的克服经典的主成分回归（CPCR）方法容易出现的多个异常点的掩盖现象；基本可以认为2008年地区经济增长与相关指标数据是匹配的，但部分地区的经济增长数据可能存在可靠性问题。

关键词: 统计数据可靠性, 稳健主成分回归, 异常值诊断

Abstract: Robust principal component regression (RPCR) methodology is a combination of robust principal component analysis and robust regression. This paper first applies RPCR and corresponding outliers detection tools to assess the reliability of Published China’s regional GDP growth rates of 2008. The results show that the RPCR methods which can better overcome the influence of outliers is more reliable than traditional methods and the problem of masking effect which exists in traditional methods can be effectively solved. The results also show that published regional GDP growth rates and selected independent indictors are mutually consistent basically, but the reliability of GDP growth rates of some areas is suspected.

Key words: The Reliability of Statistics, RPCR, Outliers Diagnostic

卢二坡张焕明. 基于稳健主成分回归的统计数据可靠性评估方法[J]. 统计研究, 2011, 28(8): 21-27.

Lu Erpo & Zhang Huanming. An Evaluation Method of Reliability of Published Statistics Data Based on Robust Principal Component Regression Method[J]. , 2011, 28(8): 21-27.