Bayesian Inference for Gaussian Copula Regression Models

Abstract: 

Bayesian Inference for Gaussian Copula Regression Models

Inference for Gaussian copula regression models is challenging when the response is discrete because computation of the true likelihood is in O(2n), where n is the sample size. The aim of this project is to develop, and evaluate by simulation, two approximations to the likelihood. These approximations can be computed in polynomial time, but to evaluate their performance the researchers will need to fit hundreds of simulated datasets, which will be quite burdensome computationally. Using hundreds of cores simultaneously, one for each dataset, will make a thorough study feasible.

Group name: 
hughesj