Vine Copulas for Multivariate Dependence
Judith Claassen | Vrije Universiteit Amsterdam | 
Overview
This presentation and python tutorial cover the use of copulas for modelling multivariate dependencies between the variables involved in compound and extreme events, organised around three areas:
Copula foundations, including Sklar’s theorem and how copulas separate marginal distributions from the dependence structure, and the main bivariate copula families used to capture symmetric, asymmetric, and tail-dependent relationships
Vine copulas for higher dimensions, including the pair-copula construction that builds multivariate models from bivariate building blocks, the tree structure of regular vines, and the C-vine and D-vine subclasses
Applications to extreme events, including random and conditional sampling from fitted copulas, and their use in characterising the joint occurrence of hazards and drivers for compound event analysis
Setup
You need Python 3.10+ and the vinecopulas package (pip is included with Python). Install the dependencies:
pip install vinecopulas notebook "pandas<2.2" "numpy<1.27" "matplotlib<3.9"Then launch Jupyter and open the notebook files:
jupyter notebook