Title: Multiple change-point for semiparametric copula models
Authors: Olivier Lopez - Sorbonne Universite Paris (France) [presenting]
Abstract: A dynamic copula model is considered. We assume that the dependence structure between some random variables is a copula belonging to a fixed parametric copula family, but with association parameter $\theta(t)$ evolving with time $t$. A multiple change-point model consists in assuming that the function $\theta(t)$ is piecewise constant, without pre-determining the times where the jumps are located. We derive finite sample bounds for maximum likelihood estimation of these times and amplitudes of jumps, and show the consistency of model selection procedures to select the appropriate number of changes.