While there is substantial need for dependence models in high dimensions, most existing models strongly suffer from the curse of dimensionality and barely balance parsimony and flexibility. In this paper, the new class of hierarchical Kendall copulas is proposed to tackle these problems. Constructed with flexible copulas specified for groups of variables in different hierarchical levels, hierarchical Kendall copulas provide a new way to model complex dependence patterns. The paper discusses properties and appropriate inference techniques for hierarchical Kendall copulas. A substantive application to German stock returns finally shows that hierarchical Kendall copulas perform very well, out-of- as well as in-sample.     «

While there is substantial need for dependence models in high dimensions, most existing models strongly suffer from the curse of dimensionality and barely balance parsimony and flexibility. In this paper, the new class of hierarchical Kendall copulas is proposed to tackle these problems. Constructed with flexible copulas specified for groups of variables in different hierarchical levels, hierarchical Kendall copulas provide a new way to model complex dependence patterns. The paper...     »