Multiple imputation is a statistical framework for analyzing incomplete data sets. The basic idea is to impute an incomplete data set several times considering the observed dependence structure, analyze each of these completed data sets separately, and then combine the results. So far, only a few exceptions in the literature focus on the generation of the imputations with vine copulas, although vine copulas are highly flexible models for multidimensional dependence. In this thesis, we propose a novel multiple imputation method based on the fully conditional specification approach using D-vine quantile regression models. We conducted a simulation study to evaluate the performance of this vine-based method and compare it to well-established multiple imputation methods. Our findings indicate that, under certain conditions, the D-vine quantile regression approach can yield enhanced performance. Furthermore, we present a real-data application in finance concerning the determination of ESG scores, where our approach demonstrates superiority over other methods.     «

Multiple imputation is a statistical framework for analyzing incomplete data sets. The basic idea is to impute an incomplete data set several times considering the observed dependence structure, analyze each of these completed data sets separately, and then combine the results. So far, only a few exceptions in the literature focus on the generation of the imputations with vine copulas, although vine copulas are highly flexible models for multidimensional dependence. In this thesis, we propose a...     »