Likelihood Inference for Life Test Data

This paper reviews the basic ideas of likelihood inference and applies them to some reliability and life testing problems connected with the exponential distribution. Various censoring schemes are considered and the likelihood approach is shown to provide a unified framework for treating exact solutions to estimation, prediction, and population-comparison problems. These solutions are provided in a well defined but non-probabilistic form. The classical s-confidence procedures depend strongly on imaginary exact repetitions of the form of the experiment being analyzed. This is often unrealistic and impractical. The data on hand can be directly analyzed by examining the likelihood function.