估计极端行为模型:分位数回归方法及其实现与应用  

Estimating Behavioral Models of Extreme Behavior: Quantile Regression Method and Its Realization and Application

作　　者：吴建南[1] 马伟[1] 

机构地区：[1]西安交通大学公共政策与管理学院,陕西西安710049

出　　处：《数理统计与管理》2006年第5期536-543,共8页Journal of Applied Statistics and Management

基　　金：国家社会科学基金项目(03CZZ006);国家自然科学基金项目(70203009)资助

摘　　要：在许多社会和管理研究中,研究者通常很感兴趣不同于期望或平均的极端行为的理论解释。这些特殊个案所包含的信息往往是研究的创新点和解决某些问题的突破口,但传统的最小平方法与最小一乘法并不适宜于这类研究问题的解决。本文讨论一种估计极端行为的理想模型:分位数回归。本文在对分位数回归的国内外研究现状进行综述后,介绍了分位数回归的模型和实现方法,并将它与最小平方法、最小一乘法进行了比较。最后探讨它在我国管理研究领域的应用方式和有关条件。In a great deal of society and management research, researchers are usually interested in theory explanation that aims to extreme behavior known from expectation or average behavior. Information including in these special cases is usually innovation of research and breach of some solutions, but traditional ordinary least squares and least absoulte deviations can not fit to solve this type of research problem. After summarizing research actuality of quantile regression inside and outside our country, this article introduces this ideal model and realization method, and compares it with OLS and LAD. Finally, this article discusses its applied mode and relative condition in management research field of our country.

关 键 词：分位数回归 最小平方法 最小一乘法 

分 类 号：O212[理学—概率论与数理统计]