B1617
Title: A correlation-shrinkage prior for the 2-dimensional Wishart model
Authors: Tomonari Sei - The University of Tokyo (Japan) [presenting]
Fumiyasu Komaki - RIKEN CBS (Japan)
Abstract: For the two-dimensional Wishart model, we propose a scale-invariant and permutation-invariant prior distribution that shrinks the correlation coefficient. The prior is characterized by a uniform distribution for Fisher's z-transformation of the correlation coefficient. The Bayesian predictive density based on our prior is shown to be minimax under the Kullback-Leibler loss.
