摘要
本文研究了线性模型中数据删除、均值漂移和方差扰动对回归系数β的形如β=(X^TX+ε)^(-1)X^TY的有偏估计影响的差异,推广与发展了文献的理论,本文认为,数据删除对上述有偏估计的影响往往大于均值漂移产生的影响,而且它们的影响分析一般能统一在方差扰动模型的影响分析之中,为此,本文讨论了方差扰动模型中基于有偏估计的Cook距离的性质。
In this paper, for the biased estimation β = (XTX + e)-1XTY of the regression coefficients, the differences of influence of some disturbance models with data-delection and variance-disturbance are discussed. The results of this paper are different from the results of [1]. It is proved that the influence of the model with da-ta-delection for βis greater than the model with mean-shift does usually. The influence analysis of the models with data-delection and mean-shift for β are unified in the influence analysis of the model with variance-disturbance. And then, we give the properties of Cook distance based on the biased estimation in the model with variance-disturbance.
出处
《应用基础与工程科学学报》
EI
CSCD
1995年第4期12-18,共7页
Journal of Basic Science and Engineering
关键词
影响分析
有偏估计
数据删除模型
均值漂移模型
方差扰动模型
COOK距离
influence analysis, biased estimation, model with data-delec-tion. model with mean-shift, model with variance-disturbance, cook distance
