左截断右删失数据下对数正态分布参数多变点的贝叶斯估计  

Bayesian estimation of parameter of lognormal distribution with multiple change points for randomly truncated and censored data

作　　者：何朝兵[1] 刘华文[2] 

机构地区：[1]安阳师范学院数学与统计学院,河南安阳455000 [2]山东大学数学学院,山东济南250100

出　　处：《福州大学学报（自然科学版）》2014年第4期507-513,共7页Journal of Fuzhou University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61174099);河南省教育厅自然科学基金资助项目(2011B110001)

摘　　要：通过添加数据得到左截断右删失数据下对数正态分布的完全数据似然函数,研究了变点位置和其它参数的满条件分布.再利用Gibbs抽样与Metropolis-Hastings算法相结合的MCMC方法得到参数的Gibbs样本,把Gibbs样本的均值作为各参数的贝叶斯估计,进行随机模拟,试验结果表明各参数贝叶斯估计的精度都较高.This paper obtains the complete - data likelihood function of lognorrnal distribution for ran- domly truncated and censored data after adding data, then studies the full conditional distributions of change -point positions and other parameters, and gets Gibbs samples of the parameters by MCMC method of Gibbs sampling together with Metropolis - Hastings algorithm, and takes the means of Gibbs samples as the Bayesian estimations of the parameters. Finally random simulation tests are conducted, and the results show that the Bayesian estimations of the narameters are fairly accurate.

关 键 词：完全数据似然函数 满条件分布 MCMC方法 GIBBS抽样 Metropolis-Hastings算法 

分 类 号：O213.2[理学—概率论与数理统计] O212.8[理学—数学]