An Inference Methodology for Life Tests With Complete Samples or Type-II Right Censoring

We present a methodology for gaining statistical inference during the actual conduct of a life test experiment that can reduce time on test and cost. The method relies on properties of conditional order statistic distributions to convert censored data into i.i.d. uniform variates. A secondary result presents a new test for uniformity based on the convolution of these i.i.d. uniform random variables that is higher in power than the benchmark Anderson-Darling statistic in certain cases.