非参数统计视角下的异方差检验设计及其应用  

A Test for Heteroscedasticity from the Perspective of Nonparametric Statistics and its Application

作　　者：杨林涛[1] 

机构地区：[1]暨南大学统计学博士后科研流动站

出　　处：《数量经济技术经济研究》2014年第11期118-131,共14页Journal of Quantitative & Technological Economics

基　　金：国家自然科学基金项目"中国战略性新兴产业的空间布局与发展路径"(71273276);教育部哲学社科青年基金项目(13YJC910004)的资助

摘　　要：以数据生成过程为导向,探讨了异方差来源的基本类型;依据非参数统计的基本思想,设计了切实可行的Mood方差检验方法与平方秩检验方法,并针对异方差来源类型,分析了相应的检验思路。蒙特卡洛模拟表明,Mood方差检验方法在异方差检验方面具有很高的检验效力;平方秩检验方法在异常经济现象情形下检验效力较低,而在其他情形下检验效力很高。同时,进一步阐释了纠正异方差的基本逻辑。Based on the Data Generated Process Oriented, the paper discusses the basic types of the source of heteroscedasticity. According to the basic idea of nonparametric statistics, it designs the feasible Mood and squared rank test methods, and it analyzes the testing sequence in view of the source of heteroscedasticity. Monte Carlo Stimulation shows that the test power of the Mood method is very high; the test power of squared rank test method is lower in the case of abnormal economic phenomenon, but it is higher in the other case. At the same time, it further explains the basic logic for the correction of heteroskedasticitv.

关 键 词：非参数统计 异方差 数据生成 蒙特卡洛 

分 类 号：F224.0[经济管理—国民经济]