一类线性模型异常数据的检验

Detection of Outliers in a Linear Model

针对来自于线性模型Y=Xβ+e的数据,给出了检验其异常数据的新方法.针对误差方差已知和未知两种情形,分别提出两个检验统计量,给出了检验该线性模型中异常数据的具体步骤和检验的拒绝域,并且证明了该检验方法的检验水平.

A new method of testing outliers in a class of linear model Y=Xβ＋e is brought forth. In the two cases of the known and unknown error variance＇, two test statistics are supplied respectively. The detailed steps i＇or testing out[iers in the linear model and the testing rejection range are both presented and the testing power of such a method proves to be effective.