BAYES-LIN: An object-oriented environment for Bayes linear local computation

Abstract

BAYES-LIN is an extension of the LISP-STAT object-oriented statistical computing environment, which adds to LISP-STAT some object prototypes appropriate for carrying out local computation via message-passing between clique-tree nodes of Bayes linear belief networks. Currently the BAYES-LIN system represents a rather low-level set of tools for a back-end computational engine, together with diagnostic graphics for understanding the effects of adjustments on the moral graph. A GUI front end, allowing interactive formulation of DAG models could be easily added, but is currently missing from the system. This document provides a very brief introduction to the system, by means of a work-through of two example computations, followed by a list of variables, functions, objects and methods provided by the system.