病态模型异常点的满秩分解法与直接非病态化法检验

Outlier Detect in Linear Model with an Abnormal Design Matrix by Full Rank Decomposition and Direct Normalization

根据病态模型删除第个数据点等价于删除其满秩分解模型的第i个数据点,然后借助于非病态模型异常点的检验方法获得病态模型异常点满秩分解法及直接非病态化法检验。

The author resolves the outlier detect linear model with an abnormal matrix by full de-composition model and direct normalization model and then using the method of outlier detect in linear model with a normal matrix, Finally examples are given with man-made data.