系统仿真学报 ›› 2016, Vol. 28 ›› Issue (5): 1070-1076.

• 仿真建模理论与方法 • 上一篇    下一篇

耿贝尔分布的最优线性矩估计

翟宇梅1, 赵瑞星2   

  1. 1.北京应用气象研究所,北京 100029;
    2.中国人民解放军61741部队,北京 100094
  • 收稿日期:2014-12-30 修回日期:2015-04-17 发布日期:2020-07-03
  • 作者简介:翟宇梅(1963-),女,河北,硕士,研究员,研究方向为天气气候分析与预测、统计建模;赵瑞星(1957-),男,山西,硕士,高工,研究方向为建模与仿真、时间序列分析。

Optimal Linear Moment Estimation Method for Gumbel Distribution

Zhai Yumei1, Zhao Ruixing2   

  1. 1. Beijing Institute of Applied Meteorology, Beijing 100029, China;
    2. Unit 61741 of PLA, Beijing 100094, China
  • Received:2014-12-30 Revised:2015-04-17 Published:2020-07-03

摘要: 线性矩法是工程实践中极值分布函数参数估计常用方法之一。为提高线性矩法对耿贝尔分布的估计精度,通过蒙特卡罗模拟比较研究了12种常用经验分布函数的估计结果,选用估计误差最小的经验分布函数构造最优线性矩估计法,并与普通矩法、线性矩法和极大似然函数法进行了比较。仿真结果表明,选择合适的经验分布函数可以提高参数估计精度;样本容量不同,最优经验分布函数不同;多数情况下最优线性矩法估计精度较高,特别是在样本容量为1 000~10 000时,其估计结果均好于普通矩法、线性矩法和极大似然函数法。

关键词: 耿贝尔分布, 蒙特卡罗模拟, 经验分布函数, 最优线性矩, 样本容量

Abstract: Linear moment is one of the important methods to estimate the extreme value distribution function in engineering practice. In order to improve the estimation performance for Gumbel distribution, the estimating results of twelve empirical distribution functions were simulated by using Monte Carlo method. The optimal linear moment estimation method was constructed based on the empirical distribution function with the minimum estimation error, and compared with ordinary moment, traditional linear moment and maximum likelihood method. The study shows that choosing an appropriate empirical distribution function can enhance the estimation accuracy. The optimal empirical distribution function varies depending on the sample size, and its result is quite satisfactory in most cases and offers the best among the other methods especially in the sample size of 1 000~10 000.

Key words: Gumbel distribution, Monte Carlo simulation, empirical distribution function, optimal linear moment, sample size

中图分类号: