异方差G-Q检验方法的改进

Improvement in Goldfeld-Quandt Heteroskedasticity Test

围绕戈德菲尔德—匡特异方差检验方法(G-Q检验)展开相关问题的研究讨论,并对该检验方法进行改进。在对线性回归模型异方差检验的两个基本问题进行分析阐述的基础上,对G-Q检验中使用的F检验方法进行辨析讨论,并在应用中针对不同形式的数据条件进行拓展;以此研究为基础,设计基于线性回归模型拟合值排序的多元线性回归模型的G-Q检验方法,并对这一方法进行统计模拟和应用检验,以证实该方法良好的检验效果和简便的使用过程。

The paper focuses on Goldfeld-Quandt heteroscedasticity test(G-Q test)and discusses some related issues including an application improvement.Based on the analysis of the two basic problems about the heteroscedasticity of the linear regression model,we made discussion and development about the F-test method in the G-Q test.Based on these studies,we designed a new G-Q heteroskedasticity test method based on the fitted values sort to test the multivariate linear regression model's heteroscedasticity.We made some statistical simulations for the related result,and it confirmed that the new test method is good and easy to use.