B0912
Title: Flexible beta regression with functional covariates: A Bayesian approach
Authors: Agnese Maria Di Brisco - University of Piemonte Orientale (Italy) [presenting]
Enea Bongiorno - Universita del Piemonte Orientale (Italy)
Aldo Goia - University of Eastern Piedmont Amedeo Avogadro (Italy)
Sonia Migliorati - University of Milano Bicocca (Italy)
Abstract: Modeling bounded continuous response variables, such as rates and proportions, is a common issue in many disciplines. Due to the constraint on the response, possible solutions are the beta and the flexible beta regression models. The latter has been recently proposed and it is based on a special mixture of betas designed to cope with (not limited to) bimodality, heavy tails, and outlying observations. These models are generalized to the case of Hilbert valued covariates. Estimation issues are dealt with through a combination of standard basis expansion and MCMC techniques. Specifically, we propose to select the most significant coefficients of the expansion through Bayesian variable selection methods that take advantage of shrinkage priors. The effectiveness of the proposal is illustrated by using numerical examples and an application of spectrometric analysis on milk specimens.
