线性回归模型中的后验影响函数

Posterior Influence Function in Linear Regression Model

本文对线性回归模型定义了后验影响函数;针对已有的先验信息的三种不同情况给出了它的分解式,讨论了其优良性;指出强影响点、异常点、高杠杆点的内在联系,讨论了上述三种点的探测方法。

In this paper, we define posterior influence function in linear regression model, decompose it into a sum of meaningful terms for three different situations, show its merits, clarify the relationship of influential observations, high leverage cases and outliers. We also discuss the detection of these points.