Optimum plan for step-stress model with progressive type-II censoring

This paper considers statistical inference on a simple step-stress accelerated life test with progressively type-II censored exponential data. In this case, the uniformly minimum variance unbiased estimators (UMVUEs) of the mean life and the failure rate at a design stress are derived. It is shown that the UMVUEs have smaller mean squared errors than the maximum likelihood estimators (MLEs). A numerical example is used to illustrate the superiority of UMVUEs to MLEs. The optimum test plans of the models are also discussed based on the variance of UMVUEs. Furthermore, we discuss the generalization of the optimal plan to more than two levels of stress.