Bayesian Logistic Regression Model Choice via Laplace-Metropolis Algorithm

Following a Bayesian statistical inference paradigm, we provide an alternative methodology for analyzing a multivariate logistic regression. We use a multivariate normal prior in the Bayesian analysis. We present a unique Bayes estimator associated with a prior which is admissible. The Bayes estimators of the coefficients of the model are obtained via MCMC methods. The proposed procedure is illustrated by analyzing a data set which has previously b"'en analyzed by various authors. It is shown that our model is more precise and computationally less taxing.