Estimation and Testing of Hypotheses for Distributions in Accelerated Life Tests.

摘要

In this paper, a procedure is presented for the interference from accelerated life tests without assuming a parametric family of failure distributions at the different stress levels. Our time transformation function is a generalization of the familiar inverse power law relationship. We consistently estimate the failure distribution at use conditions stress, and show how to test for a specified common parametric family for the life distribution under different stresses. We illustrate our procedures by considering some failure data from an industrial accelerated life test. Our results should be of interest to biometricians involved with experiments of carcinogenic substances.