A1133
Title: Parametric estimation of conditional Archimedean copula generators for censored data
Authors: Marie Michaelides - Heriot-Watt University (United Kingdom) [presenting]
Helene Cossette - Laval University (Canada)
Mathieu Pigeon - Université du Québec à Montréal (UQAM) (Canada)
Abstract: The aim is to propose a novel approach for estimating Archimedean copula generators in a conditional setting by incorporating endogenous variables directly into the generator function. Traditional copula models often rely on the simplifying assumption that the dependence structure remains fixed across all values of the covariates. The method relaxes this assumption by allowing both the strength and the shape of dependence to vary with covariates. In addition, to pinpoint the levels of a continuous risk factor where the dependence structure undergoes significant changes, an iterative splitting algorithm is introduced, which identifies optimal splitting points in the covariate space, that is, the range of possible values of a covariate. The effectiveness of the methodology is demonstrated through applications in two diverse settings, a diabetic retinopathy study and a claims reserving analysis, showing that accounting for covariate influences leads to a more accurate capture of the underlying dependence structure, thereby enhancing the applicability of copula models, for example, in medical or actuarial contexts.
