Distributions of Parameters for Uncertainty Analysis Cannot Be Defined without Using Prior Information


Hendriek C. Boshuizen, Department of Statistics and Mathematical Modelling, National Institute of Public Health and the Environment, PO Box 1, 3720 BA Bilthoven, The Netherlands. E-mail: hendriek.boshuizen@rivm.nl


Background:  Barendregt proposes a method to define an input distribution for a relative risk, as used in the probabilistic sensitivity analysis (PSA), and suggests the method is “non-Bayesian” and thus does not require prior knowledge on the probability distribution of the relative risk.

Aims:  To discuss the method from an epistemologically viewpoint.

Materials and Methods:  Examination of the underlying assumptions.

Results:  The method, like other methods to define input distributions, is Bayesian in character and the implied prior distribution is not very appealing.

Discussion:  Bootstrapping offers possibilities to be non-Bayesian, but at the price of giving only non-Bayesian answers. The method presented by Barendregt, however, can not be seen as a bootstrapping approach.

Conclusion:  Defining the distribution of a RR or any other model parameter without being a Bayesian is epistemologically impossible. This means that being explicit on prior distributions used for deriving those distributions, and justifying them, is a necessary part of suggesting new ways to define distributions.