摘要
Pena距离是研究偏态数据的一种有用工具.本文利用Pena距离研究了偏正态数据下位置回归模型的统计诊断问题,得到了位置回归模型下Pena距离的表达式,对其性质进行讨论,从而得到高杠杆异常点的判别方法. Pena距离与Cook距离、似然距离进行比较,得到在一定的条件下Pena距离优于Cook、似然距离.通过随机模拟试验研究和实例分析,表明本文提出的理论和方法是科学合理的.
Pena distance is a useful tool to study skewed data. In this paper, the statistical diagnostics of the Location regression model under skew-normal data based on the Pena distance is discussed. The expression of the Pena distance about Location regression model is obtained and its properties are discussed. Meanwhile, the discrimination of high-leverage outlier are obtained, then the conclusions obtained are that the Pena distance is better than the Cook and Likelihood distance on certain conditions. Monte Carlo simulation studies and two real examples analysis illustrate that the model and method are scienti?c and reasonable.
引文
[1]PENA D.A new statistic for influence in linear regression[J].Technometrics,2005,47(1):1-12.
[2]孟丽丽,卢志义.基于Pena距离的加权最小二乘估计的影响分析[J].数理统计与管理,2009,28(2):252-257.
[3]胡江.基于Pena距离的非线性回归模型的影响分析[J].大学数学,2012,28(5):80-85.
[4]胡江.基于Pena距离的几种回归模型的影响分析[D].南京:东南大学,2012.
[5]胡江,林金官,赵彦勇.基于Pena距离的广义线性回归模型的影响分析[J].应用数学,2017,30(3):539-546.
[6]XIE F C,LIN J G,WEI B C.Diagnostics for skew-normal nonlinear regression models with AR(1)errors.Computational Statistics and Data Analysis,2009,53:4403-4416.
[7]万文,吴刘仓,马梦蝶.偏正态数据下联合位置与尺度模型的统计诊断[J].应用数学,2017,2(9):1-10.
[8]AZZALINI A.A class of distributions which includes the normal ones[J].Scandinavian Journal of Statistics,1985,12(2):171-178.
[9]韦博成,林金官,解锋昌.统计诊断[M].北京:高等教育出版社,2009.
[10]韦博成.参数统计教程[M].北京:高等教育出版社,2006.
[11]高惠璇.实用统计方法与SAS系统[M].北京:北京大学出版社,2001.
[12]CARROLL R,RUPPERT D.Transformation and Weighting in Regression[M].New York:Wiley,1988.