Fitting Additive Copula Regression Models for Binary Outcome Regression
Additive copula regression for regression problems with binary outcome via gradient boosting [Brant, Hobæk Haff (2022); <arXiv:2208.04669>]. The fitting process includes a specialised model selection algorithm for each component, where each component is found (by greedy optimisation) among all the D-vines with only Gaussian pair-copulas of a fixed dimension, as specified by the user. When the variables and structure have been selected, the algorithm then re-fits the component where the pair-copula distributions can be different from Gaussian, if specified.