Well-tested and available software for evaluating multidimensional integrals of moderate dimensionality may be adapted for use in Bayesian inference via elementary parameter transformations. We illustrate with an example from cognitive modeling of error rates in computer-based tasks, in which the parameter being integrated is six-dimensional and the integrand itself requires a product of twenty one-dimensional integrations for each function evaluation. This method appears competitive with, and may be superior to, alternative methods when the transformations are well chosen. adaptive integration, hierarchical models, multiple integrals, posterior computation.
This article is from 'Computing Science and Statistics: Proceedings of the Symposium on Interface Critical Applications of Scientific Computing (23rd): Biology, Engineering, Medicine, Speech Held in Seattle, Washington on 21-24 April 1991,' AD-A252 938, p441-444.