A Monte Carlo method for multiple parameter estimation in the presence of uncertain data

Abstract Two characteristics that cause difficulties in formal Bayesian efforts to estimate parameters of models useful in probabilistic risk assessment are the large number of parameters involved and the presence of uncertainties in the available data. Both characteristics lead to multidimensional integrals that often cannot be evaluated using standard analytical or numerical techniques. The latter characteristic also requires the explicit development of a distribution for the uncertain data, a process that can become extremely burdensome when dealing with discrete data even when the amount of data is small. This paper discusses a Monte Carlo approach for performing the estimation that avoids these two problems. The approach is applied to problems involving discrete data and continuous data and is shown to be reasonably accurate, fast running, and easy to employ.