Evolution review of BayesCalc, a Mathematica package for doing Bayesian calculations

P. Desmedt, K. Thielemans, I.Lemahieu
pol@petdec.rug.ac.be

Abstract

Mathematica is a program for doing symbolic and numerical mathematics on computers. It is possible to extend the built-in rules of Mathematica to match the needs of a specific research field. The presented package implements the rules governing Bayesian probability theory in a Mathematica framework.

The implemented rules enable the calculation of posterior probabilities from given probabilistic relations. The main rules are the product and marginalization rule. The evolved version of BayesCalc incorporates additional features such as decision theory and hypothesis testing.

The previous version of BayesCalc dealt with symbolic calculations. Unfortunately, problems arise with some symbolic operations. Especially integrations are difficult. To overcome this problem, numerical versions of many operations were added to the package. With these numerical operations, results can still be obtained when the symbolic calculations break down.


MaxEnt 94 Abstracts / mas@mrao.cam.ac.uk