Behavior of weighted likelihood and weighted posterior methods for treatment of uncertain data

摘要

When performing a reliability assessment for a new product, the reliability engineer may be confronted with data uncertainty, i.e. situations where available data itself is subject to uncertainty. This paper discusses a few cases of data uncertainty, and proposes ways of addressing them, using the Weighted Likelihood and Weighted Posterior methods. Differences between the two methods are discussed, and further illustrated by means of a numerical example. It is shown that the way in which uncertainty is addressed affects the analysis results, and that a method should be chosen that corresponds to the type of uncertainty.