by Jukka Corander
Research Report 1999:7
Department of Statistics, Stockholm University, S-106 91 Stockholm, Sweden
Abstract
Model determination within the class of mixed graphical interaction models is considered using a Bayesian approach. The plausibility of a model is judged by the expected entropy of the corresponding predictive distribution and the relative complexity of the model. The suggested criterion becomes equivalent to the Schwarz criterion when the sample size tends to infinity. The expectation of the entropy is numerically approximated using a sampling-importance-resampling algorithm. Real data sets are used to illustrate the introduced methodology.
Key words: Bayesian
model determination, CG distribution, Entropy, Mixed graphical interaction
models, Sampling-importance-resampling,Schwarz criterion.
Last update: 2000-02-15/CE