WEIGHTED RANK REGRESSION WITH DUMMY VARIABLES FOR ANALYZING ACCELERATED LIFE TESTING DATA

Abstract

In this article, we propose a new rank regression model to extrapolate the product lifetimes at normal operation environment from accelerated testing data. Weighted least squares method is used to compensate for nonconstant error variance in the regression model. A group of dummy variables is incorporated to check model adequacy. We also developed customizing software for quick-and-easy implementation of the method so that reliability engineers can easily exploit it. Simulation studies show that, under light censoring, the proposed method performs comparatively well in predicting the lifetimes even with small sample sizes. With its computational ease and graphical presentation, the proposed method is expected to be more popular among reliability engineers.