选择重尾阈值k的Bootstrap方法  

Bootstrap Method in Selecting Heavy-tailed Threshold k

作　　者：刘维奇[1,2] 赫英迪[2,3] 邢红卫[2] 

机构地区：[1]山西大学管理科学与工程研究所,山西太原030006 [2]山西大学数学科学学院,山西太原030006 [3]广东茂名职业技术学院,广东茂名525000

出　　处：《山西大学学报（自然科学版）》2010年第4期508-512,共5页Journal of Shanxi University(Natural Science Edition)

基　　金：教育部人文社会科学研究项目(07JA63002706JA630035);山西省高校人文社科重点研究基地项目(20083006)

摘　　要：详细讨论了重尾指数估计中选取k的Sum-plot方法和Bootstrap方法,并对Hall提出的Bootstrap方法作了改进,称为M-Bootstrap方法.并利用上述三种方法对已知重尾分布进行Monte-Carlo模拟,研究它们的可行性,比较它们的稳健性,改进的M-Bootstrap方法对重尾指数的估计在某些情况下优于Bootstrap方法.We discuss the Sum-plot method and Bootstrap method in selecting k in heavy-tailed index estima- tion,and improve the Bootstrap method proposed by Hall, known as the M-Bootstrap Method. The three methods were used to study the known heavy-tailed distributions by Monte-Carlo simulation technology, in cluding their feasibility,Moreover, their robustness was compared the M-Bootstrap method was better than the Bootstrap method in some cases for heavy-tailed index estimation.

关 键 词：重尾指数 重尾阈值 Sum-plot方法 BOOTSTRAP方法 M—Bootstrap方法 

分 类 号：O212[理学—概率论与数理统计]