基于统计深度函数的G-Q检验  

The Generalization of G-Q Test Based on Statistical Depth Function

作　　者：金蛟[1] 

机构地区：[1]北京师范大学,数学科学学院,统计数据分析实验室,北京100875

出　　处：《数理统计与管理》2008年第1期100-105,共6页Journal of Applied Statistics and Management

摘　　要：G-Q检验是一种简单、有效的异方差检验方法,但该方法只适用于一个自变量,在多变量情况下,文献[1]利用主成分对样本数据进行排序,得到了G-Q检验的推广.众所周知,主成分分析是一种有效的降维方法,但其在降维的同时伴随着信息的损失,统计深度函数可作为多元数据排序的有效工具.本文基于统计深度函数得到了推广了的G-Q检验,并应用于实例。G-Q test is a simple but effective method for the test of heteroscedasticity, but it can only be used in one variable case. In multivariate situation, G-Q test is generalized in using the Principal Component Analysis method. During the course, the multivariate data can be ordered. As we all know, Principal： Component Analysis is an effective method for dimension reduction. But at the same time some information of the data is lost. Statistical depth function is effective for the arranging of multivariate data. In this paper, the G-Q test is generalized based on the statistical depth function and it is applied in one example.

关 键 词：异方差性 G-Q检验 统计深度函数 

分 类 号：O212[理学—概率论与数理统计]