岭回归分析在解决经济数据共线性问题中的应用

The Application of Ridge Regression in Solving the Problem of Collinearity of Economic Data

多重共线性是经济数据回归分析时常碰到的问题,普通最小二乘法无法很好地消除共线性对参数估计及模型预测的影响。而岭回归方法引入一个正数改进正规方程组,提供一种有偏估计的方法消除共线影响。因此,从共线性概念出发,介绍共线性问题诊断,然后基于岭回归基本思想,论述岭估计、岭参数值的确定、变量选择等问题。实例对就业相关宏观经济数据进行岭回归,通过过程与结果来说明岭回归分析可以有效处理共线性,得到较好的回归方程。

Multi-collinearity is a common problem encountered in the regression analysis of economic data,the ordinary least squares cannot eliminate the effect of collinearity on parameter estimation and prediction. But the ridge regression adds a positive number to improve the normal equations,providing a method of biased estimation to eliminate collinear effects. Therefore,based on the concept of collinearity,the problem diagnosis is introduced. Then,based on the basic idea of ridge regression,ridge estimation,ridge factor determination and variable selection are discussed. A set of employment-related macroeconomic data is exampled by using the ridge regression to resolve,through the process and results to illustrate that this method can deal with the collinear problem effectively,and get a better regression equation than ordinary least squares.