Title: Hierarchical processes for Bayesian nonparametric inference
Authors: Antonio Lijoi - University of Pavia and Collegio Carlo Alberto (Italy) [presenting]
Federico Camerlenghi - Bocconi University (Italy)
Peter Orbanz - Columbia University (United States)
Igor Pruenster - Bocconi University (Italy)
Abstract: Recent findings are discussed on the use of completely random measures for constructing priors suitable for Bayesian nonparametric inference with data that display dependence structures more general than exchangeability. The focus will be on a broad class of hierarchical processes whose distributional properties will be presented. These theoretical results are the key ingredients for devising suitable MCMC algorithms that may rely either on a ``marginal'' or on a ``conditional'' approach. Illustrations related to prediction within species sampling problems and inference on survival data will be finally displayed.