华东师范大学学报(自然科学版) ›› 2006, Vol. 2006 ›› Issue (5): 93-98.

• 数学 统计学 • 上一篇    下一篇

无失效数据的EM算法

赵海兵, 程依明   

  1. 华东师范大学 统计系, 上海 200062
  • 收稿日期:2005-04-30 修回日期:2005-09-19 发布日期:2006-09-25
  • 通讯作者: 程依明

Statistical Analysis for Zero-Failure Data with EM Algorithm (Chinese)

ZHAO Hai-bing, CHENG Yi-ming   

  1. Department of Statistics, East China Normal University, Shanghai 200062, China
  • Received:2005-04-30 Revised:2005-09-19 Published:2006-09-25

摘要： 在可靠性统计的定时截尾寿命试验中,最后一个失效时间与定时截尾时刻之间的信息常被忽略.
本文利用EM算法来处理这一情形.先用已知数据来估计出未观察到的潜在数据, 进而得出可靠性指标 的估计.并将EM算法与极大似然估计及修正极大似然估计进行了比较,可以明显看出EM方法优于后两者.最后将EM算法推广到无失效数据情形下,得出无失效数据下的指数分布平均寿命的一个估计.

关键词: 指数分布, 无失效数据, 定时截尾, EM算法, 极大似然估计, 修正极大似然估计, 指数分布, 无失效数据, 定时截尾, EM算法, 极大似然估计, 修正极大似然估计

Abstract: The paper dealt with Type I censored sample by EM algorithm. Observed failure data were used to estimate unobserved failure data, and the distribution's estimated parameter was obtained. Moreover, a comparison of EM algorithm, maximum likehood estimation and a modified maximum likehood estimation was made. Finally,EM algorithm was used to statistical analysis for zero-failure data, and to estimate the exponential distribution's parameter.

Key words: zero-failure data, time curtailed test, EM algorithm, maximum likelihood estimation, revised maximum likelihood estimation, exponential distribution, zero-failure data, time curtailed test, EM algorithm, maximum likelihood estimation, revised maximum likelihood estimation

中图分类号: