On Modeling of Failure Times Data: A Comparison of Some Well-Known Accelerated Life-Testing Models - Applications to Class-H Insulation Data

• Activity Number: 126
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Computing
• Author(s): Debaraj Sen*+ and Krishna Saha and Yogendra P. Chaubey and Sanku Dey
• Companies: Concordia University and Central Connecticut State University and Concordia University and St. Anthony's College
• Keywords: Log-normal with Box-Cox ; Inverse Gaussian ; Accelerated life test ; Weibull Models ; Gamma Models ; Bootstrap Approach

Abstract

A variety of accelerated life testing models used by many authors (see, for example, Nelson, IEEE Trans. Electrical Insulation, 1911; Chhikara & Folks, Technometrics, 1977; and Babu & Chaubey, Ann. Inst. Statist. Math., 1996) are compared. This article focuses on describing the related problem, whether from real data one may distinguish between the candidate models. To do this, we consider several aspects of some competing accelerated testing models, namely the log-normal, the log-normal with Box-Cox, the inverse Gaussian, the gamma, and the Weibull models. The maximum likelihood methods are outlined for the estimation of the parameters of these models using R. Comparison studies of these models are considered, through simulation, based on a parametric bootstrap approach of model evaluation using a Mahalanobis squared distance proposed by Allcroft and Glasbey (Statistical Modelling, 2003). Moreover, a real-life application to class-H insulation data is illustrated on the discrimination of these life testing models.