异方差模型两阶段估计的一种新方法

A new method of two-stage estimation about heteroscedastic model

异方差是线性回归模型中经常出现的问题,解决异方差问题的一个常用的方法是两阶段最小二乘法.当样本容量较小时,通过分组产生重复数据,将会损失大量样本信息,使得两阶段最小二乘法得到的估计结果不具有精确性和有效性.利用正交表将样本容量扩大,并通过分组产生重复数据,进而对数据进行两阶段估计.结果表明,该方法大大降低了估计的误差,得到了更准确的拟合模型.

Heteroscedasticity is a problem that often appears in the linear regression model,two-stage least squares method is a common method to solve this problem.When the sample size is small,grouping the samples to produce repeated data will lose a lot of information,and this will lead to the loss of accuracy and effectiveness for the estimation of two-stage least squares method.In this paper,we will expand further samples using orthogonal array,and grouping them to obtain repeated data,finally we can get the estimate by two-stage estimation.Results show that this method will reduce the error of estimation,and get a more accurate fitting model.