Modifying the Extended Neyman’s Smooth Test for Application in Accelerated Failure Time Models

Abstract

Accelerated life testing (ALT) is crucial for evaluating high-reliability units, requiring effective goodness-of-fit (GOF) techniques to test the underlying lifetime distribution across multiple stress levels. However, challenges arise due to the need to combine failure times from different stress levels to assess the adequacy of a lifetime distribution. This paper introduces a modified version of Neyman’s smooth test, called the adapted extended Neyman’s smooth test (AENST), to address these challenges within the accelerated failure time (AFT) model framework. The AENST is designed to test both Weibull and exponential distributions under constant stress with complete sampling. To evaluate its performance, the AENST is compared with the conditional probability integral transformation test (CPITT) using a simulation study. The results indicate that the AENST outperforms the CPITT in terms of power, making it a recommended tool for testing AFT models. A real dataset is also provided to demonstrate the application of the AENST.