最小二乘法回归中的一些问题

Some Problems in Least Square Estimation

最小二乘法估计通常基于：线性、独立、正态和方差齐性四个基本假定．在实际工作中，这些假定常很难满足，偏离时往往对估计结果带来影响．本文对偏离假定的情况予以讨论并给出发现和处理这些问题的方法；同时结合实例对回归中的影响点和异常点、共线性等问题予以说明．这将对我们实际工作中的数据分析具有指导性意义．

Least Square Estimation (LSE) is based on some basic assumptions, deviated from these assumptions will result in a bad estimation. In the paper we present some possible situations of not following these assumptions in data collections, findings of the problems and its solutions. At the same time, we also discussed the collinerity, diagnosis and points of influence in regression with LSE method.