Abstract: Covariance selection models were introduced by Dempster (1972). The covariance selection model with a decomposable graph is called a decomposable covariance selection model. Based on the hyper-Markov property (Dawid and Lauritzen (1993)), the exact distribution of the Maximum Likelihood Estimator (MLE) of the concentration matrix in the decomposable covariance selection model is given.
Key words and phrases: Covariance selection model, decomposable graph, inverted Wishart distribution, multivariate normal model, Wishart distribution.