B0222
Title: Imprecise statistical inference based on the log-rank test for step stress accelerated life testing data
Authors: Sultan Albalwy - Durham University (United Kingdom) [presenting]
Frank Coolen - Durham University (UK)
Jonathan Cumming - Durham University (United Kingdom)
Abstract: A new imprecise nonparametric statistical method is developed for step stress accelerated life testing data, where the Arrhenius link function is implemented for the data analysis. This function expresses the relationship between the lifetime and the applied stresses in terms of temperature to link the scale parameters of different stress levels. This method consists of three steps. First, it transforms failure times that occurred by different strategies of experimental settings at higher stress levels to the normal stress level. Second, it creates imprecision based on the log-rank test on the accelerating parameter for which the null hypothesis, that all failure times come from the same distribution, is not rejected. This imprecision allows for the transformation of failure times into interval values at the normal stress level, where it is assumed that these transformed failure times are not distinguishable from failure times occurring at the normal stress level. Third, nonparametric predictive inference is applied to the transformed data to provide robust predictive inference. This method leads to more imprecision if data are used from higher stress levels or in case of model misspecification. The performance of this method is evaluated by simulation studies.