B0546
Title: Orthogonal decomposition of probability densities in Bayes spaces
Authors: Christian Genest - McGill University (Canada) [presenting]
Abstract: Bayes spaces were initially designed to provide a geometric framework for modelling and analyzing distributional data. It recently came to light that this methodology yields a novel orthogonal decomposition of bivariate probability distributions into an independent and an interaction part. New insights into this result will be offered by reformulating it using Hilbert space theory, and a multivariate extension will be developed using a distributional analogue of the Hoeffding-Sobol identity. A connection between the resulting decomposition of a multivariate density and its copula-based representation will also be highlighted. The approach will be illustrated with geochemical data.
