摘要
提出多元正态分布整体推断方法。通过仿真结果与试验数据有机结合,该方法能够将多个不同状态下的多变量(多指标)的试验数据作为一个整体进行统计分析,给出多元正态分布均值和协方差矩阵的整体估计。文中详细讨论了多个状态下二元正态分布均值和方差的无偏估计和置信区间估计。与传统方法相比,该方法不但具有更高的精度,而且解决了一种状态下只有一个试验数据时多变量(多指标)产品可靠性评定的难题。此外,文中还给出一种均值和方差仿真结果的极小子样检验方法。
A integral inference method for multivariate normal distributions is presented. Combining the simulation results and the experiment data, it can totally infer the experiment data of the multivariate normal distributions in multi-state. Then it can establish the estimators of means and covariance matrix of the multivariate normal distribution ,and estimate their confidence limits and intervals in each state. Compared with traditional methods, this method not only has higher precision but also solves the problem of reliability assessment with only one experiment da- tum of the multivariate normal distribution in a state. Besides,a method for testing simulation resuits of means and variances in multi-state is also presented with very small sample.
出处
《航空动力学报》
EI
CAS
CSCD
北大核心
2005年第6期905-909,共5页
Journal of Aerospace Power
基金
北京市重点学科工程力学项目(XK100060418)
关键词
航空
航天推进系统
多元正态分布
小子样
整体估计
可靠性
仿真
aerospace propulsion system
integral estimate
reliability multivariate normal distribution
small sample
simulation
