We propose a Bayesian procedure to estimate heteroscedastic variances of the regression error term, when the form of heteroscedasticity is unknown. We use prior information that is elicited from the well-known Eicker-White Heteroscedasticity Consistent Variance- CovarianceMatrix Estimator, and then useMarkov ChainMonte Carlo algorithm to simulate posterior pdf's of the unknown heteroscedastic variances. In addition to numerical examples, we present an empirical investigation of the stock prices of Japanese pharmaceutical and biomedical companies.
Copyright (C) 2003-2007 by Institute of Economic
Research.All rights reserved.
