Reliability evaluation for Weibull distribution under multiply type-I censoring
来源期刊：中南大学学报(英文版)2015年第9期
论文作者：JIA Xiang JIANG Ping GUO Bo
文章页码：3506 - 3511
关键词：multiply type-I censoring; generalized confidence interval; maximum likelihood estimation; least-squares estimation
Key words：multiply type-I censoring; generalized confidence interval; maximum likelihood estimation; least-squares estimation
Abstract: The multiply type-I censoring represented that all units in life test were terminated at different times. For estimations of Weibull parameters, it was easy to compute the maximum likelihood estimation (MLE) and least-squares estimation (LSE) while it was hard to build confidence intervals (CI). The concept of generalized confidence interval (GCI) was introduced to build CIs of parameters under multiply type-I censoring. Further, GCI based on LSE and GCI based on MLE were proposed. It is mathematically proved that the former is exact and the latter is approximate. Besides, a Monte Carlo simulation study and an illustrative example also turn out that the GCI method based on LSE yields rather satisfactory results by comparison with the ones based on MLE. It should be clear that the GCI method is a sensible choice to evaluate reliability under multiply type-I censoring.
JIA Xiang(贾祥), JIANG Ping(蒋平), GUO Bo(郭波)
(College of Information Systems and Management,
National University of Defense Technology, Changsha 410073, China)
Abstract:The multiply type-I censoring represented that all units in life test were terminated at different times. For estimations of Weibull parameters, it was easy to compute the maximum likelihood estimation (MLE) and least-squares estimation (LSE) while it was hard to build confidence intervals (CI). The concept of generalized confidence interval (GCI) was introduced to build CIs of parameters under multiply type-I censoring. Further, GCI based on LSE and GCI based on MLE were proposed. It is mathematically proved that the former is exact and the latter is approximate. Besides, a Monte Carlo simulation study and an illustrative example also turn out that the GCI method based on LSE yields rather satisfactory results by comparison with the ones based on MLE. It should be clear that the GCI method is a sensible choice to evaluate reliability under multiply type-I censoring.
Key words:multiply type-I censoring; generalized confidence interval; maximum likelihood estimation; least-squares estimation
