Vine Copulas are popular dependence models that provide flexible multivariate distribution classes by representing a joint distribution as univariate margins plus bivariate copulas characterizing the dependence structures. Sometimes, one is interested in sampling conditional values from those vine copula distributions. If, for any given R-vine, all required components for the conditional distribution are given directly in the representation of the vine, the conditional density can be determined easily. Other conditional densities, however, cannot be expressed because parts of the formula are not given and thus direct sampling from such conditional distribution can be hard. A feasible way to sample from conditional distributions like this is to use a Markov chain Monte Carlo (MCMC) approach, concretely an extension of the Hamiltonian Monte Carlo (HMC) algorithm – the No-U-Turn Sampler that is implemented in the probabilistic programming language Stan. By using that, we take advantage of the need for only proportional densities. By performing various simulation setups, we test whether the sampler proposed by M.Sc. Ariane Hanebeck is correctly sampling from any conditional vine copula distribution. Moreover, we apply the proposed sampler for the analysis of the Uranium data set.     «

Vine Copulas are popular dependence models that provide flexible multivariate distribution classes by representing a joint distribution as univariate margins plus bivariate copulas characterizing the dependence structures. Sometimes, one is interested in sampling conditional values from those vine copula distributions. If, for any given R-vine, all required components for the conditional distribution are given directly in the representation of the vine, the conditional density can be determined...     »