剔除相关性的最小二乘研究

本文提出了一种新的回归模型,剔除相关性的最小二乘,它有效的克服了变量间的相关性,兼顾到变量的筛选。并与最小二乘、向后删除变量法、偏最小二乘比较分析。发现剔除相关性的最小二乘能很好的处理自变量间多重相关性,对变量进行有效的筛选,克服了回归系数反常的现象。

The Least Squares Removed Correlation

This paper presents a new linear regress model,the least squares removed correlation,it will effectively overcome the correlation between variables and take account of the variable selection.then comparing with least squares,back to delete variables,partial least squares.the conclusion shows this article's model can do well with the correlation between multiple variables and selects variables effectively. Overcome the abnormal phenomenon of regression coefficients.