MODELING BEHAVIORAL SYNDROMES USING BAYESIAN NETWORKS

In this paper Bayesian networks modelling is applied to a multidimensi onal model of depression. The characterization of the probabilistic mo del exploits expert knowledge to associate latent concentrations of ne urotransmitters and symptoms. An evolution perspective is also conside red. Specific criteria are introduced to detect the influence of the l atent variable on the observation of symptoms. The Bayesian analysis i s carried out using Gibbs sampling technique which is implemented in t he BUGS software. The estimation phase leads to the selection of sympt oms entering into the definition of behavioral syndromes. Results on r eal data are discussed. The last section deals with simulation experim ents. Simulation results confirm our methodological choices. Results o f the paper can enlarge to the central problem of the management of la tent variables in Bayesian networks modelling. (C) 1998 Elsevier Scien ce B.V. All rights reserved.