Nonparametric Estimation and Goodness-of-Fit Testing of Hypotheses for Distributions in Accelerated Life Testing

This paper presents a nonparametric approach to accelerated life testing by deleting the requirement that the common parametric family of life distributions under all the stresses be specified in advance. The requirement that the time transformation function be specified is retained, and a version of the familiar inverse power law is considered. A s-consistent estimate of the failure distribution at use stress, and a test of the hypothesis that the underlying failure distribution belongs to a specified family are given. Approximate s-confidence bounds for the failure distribution at use stress are obtained. The approach is illustrated in an example using real data.