References & Citations
Statistics > Methodology
Title: Extending one-factor copulas
(Submitted on 8 Dec 2016)
Abstract: So far, one-factor copulas induce conditional independence with respect to a latent factor. In this paper, we extend one-factor copulas to conditionally dependent models. This is achieved through new representations which allow to build new parametric factor copulas with a varying conditional dependence structure. We discuss estimation and properties of these representations. In order to distinguish between conditionally independent and conditionally dependent factor copulas, we provide a novel statistical test which does not assume any parametric form for the conditional dependence structure. Illustrations of our framework are provided through examples, numerical experiments, as well as a real data analysis.
Submission history
From: Nathan Uyttendaele [view email][v1] Thu, 8 Dec 2016 21:36:10 GMT (213kb,D)
Link back to: arXiv, form interface, contact.