Bootstrap方法在导弹落点密集度评估中的误用  

Misapplication of Bootstrap method in dispersion assessment of missile fall points

作　　者：郑小兵[1,2] 董景新[1] 孟令晶 张志国 

机构地区：[1]清华大学精密仪器与机械学系精密测试技术及仪器国家重点实验室 [2]大连91550部队

出　　处：《中国惯性技术学报》2012年第2期239-243,共5页Journal of Chinese Inertial Technology

基　　金：国家"十一五"预研项目(51308050208)

摘　　要：以导弹落点密集度评估作为研究基础,对利用Bootstrap方法提升小子样下落点密集度估计精度问题的自助法进行分析。在简要描述Bootstrap方法数学思想和应用方法的基础上,从理论分析和仿真验证两个方面证实了参数Bootstrap方法生成的自助统计量围绕经典估计值呈现正态分布,对于经典估计结果没有精度的提高;并且推导了自助统计量分布与估计偏差分布之间的关系,指出在小子样条件下参数Bootstrap方法自助统计量在分布近似上存在偏差;最后本文论证了随机加权法应用于密集度估计同样不能取得明显的改进效果。Based on the assessment of missile＇s impact point dispersion, the Bootstrap method was applied to promote the evaluation accuracy of dispersion in small-sample test. The mathematic thought and application strategy of Bootstrap was introduced. Then theoretical and experimental investigations were implemented which proved that Bootstrap statistics created by parametric Bootstrap were distributed in a nearly normal fashion around classic evaluation and that estimation precision was not improved comparing with classic statistical methods. Based on analysis about the relationship between Bootstrap statistics distribution and estimation error distribution, it was proposed that in small sample case the difference of their approximation could not be ignored. At last it was proved that the application of random weighting method in dispersion assessment could not improve the accuracy similarly.

关 键 词：落点密集度 BOOTSTRAP方法 参数Bootstrap方法 随机加权法 

分 类 号：U666.1[交通运输工程—船舶及航道工程]