异方差性的诊断方法及数据属性影响的实证分析

Diagnosis Method of Heteroscedasticity Testing and Analysis of Data Essence Influence

本文对异方差性检验常用方法作了总结评价,从数据角度来讨论数据属性对异方差性检验的影响.利用Breusch-Pagan检验、和White检验,通过删除异常点,影响点前后所作模型的异方差检验,得到完全相反的结论.因此,这些发现表明,作异方差检验时要十分关注异常数据影响诊断,这对我们进行经济建模,作模型诊断有一定的指导作用.

In this article, we summarized and evaluated the methods that were often utilized to test Heteroscedasticity. We discussed that Heteroscedasticity Testing was affected by data essence from datum .We got an inverse conclusion totally with Breusch-Pagan Testing and White Testing ,throng which we made Heteroscedasticity Testing after and before we deleted outliers and influence points. Thereby, These findings have demonstrated that when we make Heteroscedasticity Testing we should quite pay attention to diagnosis of outliers and influence points. There should be certain instruction to build models and models diagnosis.