In many cases, the dependence structure between random variables varies according to the values of measured covariates. A natural approach studying these type of models is the usage of conditional copulas. Here we provide an introduction to this concept and propose a novel fully nonparametric method for the estimation of conditional copula densities. Our procedure is based on transformation kernels combined with local linear regression. The asymptotic properties are studied and a bandwidth selection procedure is suggested. We analyze the performance of this estimator in a simulation study containing a variety of scenarios, using univariate as well as multivariate covariates. In all scenarios, the method proved to work well and improves upon an estimator that does not take the covariates into account.     «

In many cases, the dependence structure between random variables varies according to the values of measured covariates. A natural approach studying these type of models is the usage of conditional copulas. Here we provide an introduction to this concept and propose a novel fully nonparametric method for the estimation of conditional copula densities. Our procedure is based on transformation kernels combined with local linear regression. The asymptotic properties are studied and a bandwidth selec...     »