B0284
Title: Stationary vine copula models for count data
Authors: Thomas Nagler - LMU Munich (Germany) [presenting]
Abstract: Stationary vine copula models for joint modelling of the cross-sectional and serial dependence in multivariate time series are introduced. The existing framework is extended to handle discrete and mixed-type random variables, presenting a pair-copula decomposition formula for generalized densities in the discrete case. The algorithm allows for efficient evaluation of the log-likelihood function of a regular vine specification with mixed margins. A stepwise maximum-likelihood approach is proposed for sequential parameter estimation, proving the asymptotic normality and consistency of the estimates. Through simulations and real-world data on daily sales, the model's performance is validated and its practical applicability is demonstrated.
