Objective Bayesian reference analysis for system reliability

Abstract

The system reliability problem is important for many areas, as medicine and engineering. There are many ways to do estimation of system reliability and your components. In general, the presented solutions are extremely restricted and its application is given only in very particular cases. The main studies in reliability were lead in the theory of competitive risks. The proposal of the project is to develop a method of distribution (survival) estimation of series systems (competitive risks) or parallel systems. The perspective used in this work will be the Bayesian ones. By the difficulty to give the prior distributions for the amounts of interest, we will work with the method developed by Berger and Bernardo (1989). This method describes one way to build a non-informative prior (reference prior). In this in case, the results produced by estimation process had only depended on the assumed model and the observed data. Applications of this theory also appear in finances and analysis of risk. (AU)