C. Ausin Olivera, M. Kalli
We propose a novel bivariate copula model able to capture both central and tail dependence of the joint distribution. We use a Bayesian nonparametric approach and introduce a random copula model based on infinite partitions of unity. We define a hierarchical prior over an infinite partition of the unit hypercube which has a stick-breaking representation leading to an infinite mixture of products of independent betas. We capitalise on the stick-breaking representation and implement a Gibbs sampler to proceed to inference and sample from the posterior. Our empirical analysis includes both simulated and real data and compares the out-of-sample predictive performance of our bivariate copula model to popular bivariate copulas (e.g. Joe, Claton) used to capture tail dependence. For both real data applications (insurance and financial returns) our approach outperforms the competitive models.
Keywords: Bayesian nonparametrics, Copulas, Dirichlet Process Mixtures, Slice sampling, Tail Dependence
Scheduled
Bayesian Inference for Economic Models
June 11, 2025 10:30 AM
Sala de prensa (MR 13)