Conditional Inference for Generalized Life Model Parameters

The conditional inference has been applied to distributions belonging to the location-scale family or those that can be converted to this family. In this paper, we will give a new application for this approach to cover the situation where the distributions do not belong to the location-scale family by converting them to a generalized life model with scale and shape parameters. For measuring the performance of this approach compared to unconditional inference, the coverage rates and the average lengths of the intervals have been obtained via Monte Carlo simulations. The simulation results indicated that the conditional intervals possess good statistical properties and can perform quite well even when the sample size is extremely small compared to the classical intervals based on the asymptotic maximum likelihood estimates (AMLEs).Finally, a numerical example is given to illustrate the confidence intervals developed in this paper.