Robust Bayesian Analysis, an Attempt to Improve Bayesian Sequencing

Abstract

Bayesian sequencing of radiocarbon dates deals with the problem that in most cases there does not exist an unambiguous way to define the so-called prior function, which represents information in addition to the result of the 14C measurements alone. However, a random choice of a particular prior function can lead to biased results. In this paper, "robust Bayesian analysis," which uses a whole set of prior functions, is introduced as a more reliable method. The most important aspects of the mathematical foundation and of the practical realization of the method are described. As a general result, robust Bayesian analysis leads to higher accuracy, but paid for with reduced precision. Our investigations indicate that it seems possible to establish robust analysis for practical applications.