摘要
针对线性回归模型中协方差阵扰动对Stein岭型主成分估计β(P)G的影响问题进行研究.证明了β(P)G的某种极限是数据删除模型的Stein岭型主成分估计;建立了β(P)G与G-M模型的Stein岭型主成分估计β(P)之间的关系;定义了度量扰动影响的距离测度DG,并给出了DG的多种计算式;最后通过实例验证其有效性.
In this paper,the issue of influence analysis of covariance matrix disturbance on stein ridge type principal components estimator(SRPCE)in linear regression model is studied.We prove that in the data deletion model,some limit of SRPCE which in the regression model with covariance matrix disturbance is SRPCE.Then,we set up the relationships among β(P)Gand β(P).Next,we define the distance measure DG,which can be assessed the disturbing influence.Afterwards,we give several calculation formulas of DG.Finally,apractical example is presented to illustrate the effectiveness of this method.
引文
[1]王松桂.线性统计模型[M].北京:高等教育出版社,1999:34-40.
[2]韦博成,鲁国斌,史建清.统计诊断引论[M].南京:东南大学出版社,1991:38-78.
[3]朱秀娟,李宁.协方差阵扰动模型的影响分析[J].应用概率统计,1993(4):356-370.
[4]林路.协方差阵扰动模型岭估计的影响分析[J].工程数学学报,1995,12(3):83-88.
[5]Zhang S L,Qin H.Influence analysis of covariance matrix disturbance in a linear model with respect to a restricting condition[J].Acta Mathematica Scientia,2006,26(26):621-628.
[6]Zhang S L,Qin H.Influence analysis on a ridge estimator in a multivariate regression model with covariance matrix disturbance[J].J Math,2010(1):157-162.
[7]杨莲.几类统计模型的局部影响分析研究[D].重庆:重庆大学,2014.
[8]赵帅.带约束线性模型LIU估计的影响分析[D].北京:北京交通大学,2015.
[9]朱宁,李建军,李兵.一种有偏岭-压缩组合估计的新形式[C]//第八届中国青年运筹信息管理学者大会论文集.桂林:[出版者不详],2006:287-290.
[10]朱宁,刘庆华,周桂兰,等.均方误差准则下的几乎无偏Stein岭型主成分估计的优良性[J].河南师范大学学报(自然科学版),2017,45(5):1-6.
[11]朱宁,严冠东.Stein岭型主成分估计下的单个数据删除模型的研究[J].统计与决策,2015(14):16-18.
[12]朱宁,严冠东,刘庆华.Stein岭型主成分估计下多个数据删除模型的强影响分析[J].汕头大学学报(自然科学版),2015,30(2):20-27.
[13]王松桂.线性模型的理论及其应用[M].合肥:安徽教育出版社,1987:165-172.
[14] Cook R D.Detection of Influential Observation in Linear Regression[J].Technometrics,1977,42(1):65-68.
[15]王松桂.线性回归诊断(I)[J].数理统计与管理,1985(6):38-49.
[16]同济大学数学系.工程数学线性代数[M].北京:高等教育出版社,2014:29-33.