Dependence Model Assessment and Selection with DecoupleNets

Abstract

Neural networks are suggested for learning a map from d-dimensional samples with any underlying dependence structure to multivariate uniformity in $d^{\prime }$ dimensions. This map, termed DecoupleNet, is used for dependence model assessment and selection. If the data-generating dependence model was known, and if it was among the few analytically tractable ones, one such transformation for $d^{\prime }=d$ is Rosenblatt’s transform. DecoupleNets have multiple advantages. For example, they only require an available sample and are applicable to $d^{\prime }<d$, in particular $d^{\prime }=2$. This allows for simpler model assessment and selection, both numerically and, because $d^{\prime }=2$, especially graphically. A graphical assessment method has the advantage of being able to identify why, or in which region of the domain, a candidate model does not provide an adequate fit, thus, leading to model selection in particular regions of interest or improved model building strategies in such regions. Through simulation studies with data from various copulas, the feasibility and validity of this novel DecoupleNet approach is demonstrated. Applications to real world data illustrate its usefulness for model assessment and selection. Supplementary materials for this article are available online.

Keywords:

• Copulas
• Graphical approach
• Model assessment
• Model selection
• Neural networks
• Rosenblatt transformation

Supplementary Materials

All types of results can be reproduced with the code accompanying this publication and the R package gnn in its current latest version 0.0-4.

Disclosure Statement

The authors report there are no competing interests to declare.

Additional information

Funding

The author acknowledges support from NSERC (RGPIN-2016-03876).