Characterizing Asymptotic Dependence via Conditional Kendall’s tau
Alexandru V. Asimit & Russell Gerrard – Cass Business School,
Yanxi Hou – Georgia Institute of Technology and Liang Peng – Georgia State University
Analysing the concomitant extreme events has become of vast interest in many applications. The first step in statistical extreme inference is to make a binary decision for the presence of asymptotic dependence or independence. Therefore, distinguishing between (only) these possible scenarios represents a crucial piece of information before starting any forecasting. Since the strength of dependence can be characterised via some measures of association, the Kendall’s tau is a good candidate to simply identify asymptotic dependence or independence. The main advantage of using this measure is due to its simplicity and ease of communication beyond the statistical academic community and practitioners. Convincing applications are detailed within the environmental statistics, insurance and finance fields. At a quantitative level, a non-parametric estimator is proposed in order to make the appropriate decisions before making any predictions.
