Gaussian Process Single Index Models for Conditional Copulas

TL;DR

This paper introduces a Bayesian nonparametric approach using Gaussian processes for modeling conditional copulas with covariates, enabling flexible inference and model selection in complex dependence structures.

Contribution

It proposes a novel Bayesian framework with sparse Gaussian process priors for flexible calibration functions in conditional copulas, including a new model selection criterion and permutation-based assessment.

Findings

Effective joint estimation of marginal and calibration functions.

Demonstrated superior performance in simulations with various copulas.

Successful application to real-world wine data.

Abstract

Parametric conditional copula models allow the copula parameters to vary with a set of covariates according to an unknown calibration function. Flexible Bayesian inference for the calibration function of a bivariate conditional copula is proposed via a sparse Gaussian process (GP) prior distribution over the set of smooth calibration functions for the single index model (SIM). The estimation of parameters from the marginal distributions and the calibration function is done jointly via Markov Chain Monte Carlo sampling from the full posterior distribution. A new Conditional Cross Validated Pseudo-Marginal (CCVML) criterion is introduced in order to perform copula selection and is modified using a permutation-based procedure to assess data support for the simplifying assumption. The performance of the estimation method and model selection criteria is studied via a series of simulations using correct and misspecified models with Clayton, Frank and Gaussian copulas and a numerical application involving red wine features.