Posterior consistency of dirichlet location-scale mixture of normals in density estimation and regression

We provide sufficient conditions under which a Dirichlet location-scale mixture of normal prior achieves weak and strong posterior consistency at a true density. Our conditions involve both the prior and the true density from which observations are obtained. We consider it to be a significant improvement over the existing results since our conditions cover the case of fat tailed densities like the Cauchy, with a standard choice for the base measure of the Dirichlet process. This provides a wider choice for using these popular mixture priors for nonparametric density estimation and semiparametric regression problems. © 2006, Indian Statistical Institute.

Duke Scholars

Author Surya Tapas Tokdar Statistical Science