Adapting Subregion-Adaptive Integration Software to Problems in Bayesian Inference

Alaattin Erkanli, Robert E. Kass, Alan Genz
Duke University, Carnegie Mellon University, Washington State University

Nov 30 1991

This paper discusses software, currently under development, for application of subregion-adaptive numerical integration methods to multiparameter Bayesian inference problems. Following Genz and Kass (Proceedings of the 23rd Interface), a parameter transformation is used to transform the domain of integration to a multidimensional unit cube. On the cube, well-tested software developed by Genz and colleagues may be used. We consider (i) the kinds of problems this software should be able to solve, and the inputs to it required of the user, (ii) alternative transformations to the unit cube, and (iii) progress in implementation.


Gaussian quadrature, multiple integrals, posterior calculations


PDF icon 1992-12.pdf