Computation on Bayesian Graphical Models
David J. Spiegelhalter, Andrew C. Thomas, Nicola G. Best · 1996
Abstract We describe the graphical expression of conditional independence assumptions and associated factorisations of the joint distribution of both observed and unobserved quantities. Such factorisations encourage local computational techniques, and a language and program for Gibbs sampling is discussed with examples concerning measurement error and spatial smoothing of cancer rates.