Réunion d'été SMC 2023

Ottawa, 2 - 5 juin 2023

Prix Krieger-Nelson

DR. JOHANNA G. NEŠLEHOVÁ, McGill Extreme or not extreme: the intricacies and challenges of rare event modeling [PDF]: Accounting for dependence between rare events such as severe storms, floods, or large financial losses is essential for sound risk management. This can be accomplished very effectively by combining extreme-value theory with copula modeling, and by relying on the component-wise ranks of multivariate data to make inference. I will first explain how copulas can adequately capture the dependence between extreme events when the commonly used asymptotic extreme-value models are unsuitable. As extending these ideas to a large collection of risks calls for hierarchical models, I will next show how their structure might be inferred by identifying patterns in large-scale rank correlation matrices. Finally, I will use some of these techniques to explore the behavior of the largest claim size in a portfolio of possibly dependent insurance claims.