首页 | 本学科首页   官方微博 | 高级检索  
     检索      


Bayesian Methods for a Growth-Curve Degradation Model with Repeated Measures
Authors:Robinson  Michael E  Crowder  Martin J
Institution:(1) Department of Mathematics and Statistics, University of Surrey, Guildford, GU2 5XH Surrey, U.K
Abstract:The increasingreliability of some manufactured products has led to fewer observedfailures in reliability testing. Thus, useful inference on thedistribution of failure times is often not possible using traditionalsurvival analysis methods. Partly as a result of this difficulty,there has been increasing interest in inference from degradationmeasurements made on products prior to failure. In the degradationliterature inference is commonly based on large-sample theoryand, if the degradation path model is nonlinear, their implementationcan be complicated by the need for approximations. In this paperwe review existing methods and then describe a fully Bayesianapproach which allows approximation-free inference. We focuson predicting the failure time distribution of both future unitsand those that are currently under test. The methods are illustratedusing fatigue crack growth data.
Keywords:degradation data  failure time distributions  likelihood methods  Markov chain Monte Carlo  nonlinear growth curves  predictive distributions  random effects  survival analysis  two-stage regression
本文献已被 PubMed SpringerLink 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号