Abstract
We present a copula-based imputation method for multivariate missing data withgeneric patterns and complex dependence structure. The method is based on theconditional density functions of the missing variables given the observed ones. Thesefunctions can be derived analytically once parametric models for the margins andthe copula are specified. When analytical derivations are not feasible, the marginsare estimated non-parametrically through local likelihood methods. We describeboth the analytic and the semiparametric version of the the copula-based imputa-tion method and investigate their performance in terms of preservation of both thedependence structure and the microdata through Monte Carlo studies. Moreover,we present a comparison with classical imputation methods. We have implementedand made available the method through the R packageCoImp. We provide an illus-tration of how to handle the imputation through the R package, i.e. a descriptionof its main functions, their output and usage on real data sets.