Estimation of High-Dimensional Copula-Based CUB Models Using Composite Likelihood

Multivariate extensions of CUB (Combination of discrete Uniform and shifted Binomial) models via copulas face severe computational challenges due to the high number of terms required to evaluate the joint probability mass function. In this study, we propose a pairwise likelihood approach to overcome these limitations by focusing on bivariate marginal distributions. The methodology involves a two-step estimation procedure: individual bivariate models are fitted for each pair, and the resulting estimates are then summarized to obtain the parameters of the full joint model. Simulation results using Archimedean copulas demonstrate that the method is computationally efficient and provides accurate parameter estimates.