基于右截尾数据的近极值事件的态密度估计  

Estimating Density of Near-Extreme Events for Right-Censored Data

作　　者：朱建国[1] 陈果[2] 黄超[2] 

机构地区：[1]南京工业职业技术学院数理学院,江苏南京210046 [2]东南大学数学系,江苏南京210096

出　　处：《南京师大学报（自然科学版）》2010年第4期33-38,共6页Journal of Nanjing Normal University(Natural Science Edition)

摘　　要：右截尾数据在实际数据中经常出现,如材料的疲劳试验等.本文研究基于右截尾数据的近极值事件的态密度(DOS)估计问题.首先定义右截尾数据类型下的态密度,接着推导了平均态密度的精确表达式.然后在大样本情形,利用蒙特卡洛(Monte-Carlo)方法、核密度估计方法等来估计平均态密度.最后应用Monte-Carlo模拟和亚马尔半岛的夏天气温数据说明了本文的方法,结果表明平均态密度的近似结果与经验结果均拟合得很好.Right-censored data often occur in the actual data, such as the fatigue test and so on. This paper studies the estimation of density of states（DOS） for right-censored data. We first redefine the DOS and the exact expression of mean DOS is derived. We show the approximate forms of the mean DOS by the theory of extreme value statistics and domains of attraction under the large-sample cases. Finally, Some Monte Carlo simulations and Yamal Peninsula summer temperature data are been used to study the density of near-extreme events, which are found to agree with the theoretical results.

关 键 词：右截尾数据 态密度 极值统计 吸引场 蒙特卡洛方法 核密度估计 

分 类 号：O212[理学—概率论与数理统计]