一种基于MCMC的变点模型在可靠性数据分析中的应用  

Application of Change Point Model Based on the MCMC Method in the Analysis of Reliability Data

作　　者：林静[1] 韩玉启[1] 朱慧明[2] 

机构地区：[1]南京理工大学,南京210094 [2]湖南大学,长沙410079

出　　处：《中国机械工程》2006年第14期1451-1455,共5页China Mechanical Engineering

基　　金：新世纪优秀人才支持计划资助项目(NCET);湖南省自然科学基金资助项目(05JJ0130)

摘　　要：引入一种广泛应用于临床试验的贝叶斯变点模型,从随机过程的角度,利用时间滞后函数对系统可靠性数据进行变点分析,进而更准确地评估出伴随变量对系统寿命分布的影响;运用基于Gibbs抽样的MCMC方法动态模拟出参数后验分布的马尔可夫链,给出随机截尾条件下参数的贝叶斯估计;利用BUGS软件包进行建模仿真分析的结果,证明了模型在可靠性应用中的直观性与有效性。This paper brought forward a Bayesian change point model, which was popular in clinical trials. It used the time lagged regression function based on the stochastic process to do the change point analysis for the system reliability data. By that it evaluated the covariates＇ influence on the system life distribution more correctly. The MCMC method based on Gibbs sampling was used to simulate the Markov Chain of the parameters＇ posterior distribution dynamically. And then, it gives out the parameters＇ Bayesian estimation under the conditions of the random truncated test. Finally, it utilizes the results of the data simulation to prove the objectivity and validity of the model by using the BUGS package.

关 键 词：贝叶斯分析 可靠性 MCMC 变点分析 

分 类 号：O212[理学—概率论与数理统计]