Le lundi 14 avril 2014 à 15:00 - UM2 - Bât 09 - Salle de conférence (1er étage)Radu Craiu
The conditional copula device allows a wider range for using copula models in the study of dependence between response variables in regression settings. We consider situations in which the dependence structure varies with covariates and we introduce a Bayesian model which allows for joint estimation of the marginal and joint models. The dependence between the copula parameter and the covariates is modelled via polynomial splines. We discuss the computation algorithm required for inference and demonstrate the performance of model selection techniques on simulated and real data.