Mini-Courses
Mini-Courses in Causality and Extremes
February 12 - 13, 2024
Venue: Lectures will be in room Uni Mail MR060; see here for an access description.
Program:
3 x 1.5h mini-course on extremes taught by Chen Zhou (Erasmus University Rotterdam)
3 x 1.5h mini-course on causality taught by Nicola Gnecco (University of California, Berkeley)
The career session will be a discussion with panelists. We invite young researchers to send questions or discussion topics of interest to gsem-causality-extremes@unige.ch.
Chen Zhou is Professor of Mathematical Statistics and Risk Management at the Econometrics Institute, Erasmus University Rotterdam. Chen’s research focuses on extreme value statistics and its applications in quantitative risk management, financial stability and financial regulation. He serves as the area editor of Extremes in Economics, Finance and Insurance.
Extreme Value Statistics: from classics to recent advances
Abstract: This mini-course covers topics ranging from classic Extreme Value Statistics to recent research related to graphical models. Univariate extreme value statistics provide statistical methods to conduct inference regarding the tail region of a univariate random variable, such as high quantiles and exceeding probabilities. In the context of multivariate extreme value statistics, the focal point turns to infer dependence across tail events described by multiple random variables, namely tail dependence. In recent advances, the number of underlying random variables can be of high dimension, which poses further challenges in statistical methods to uncover the tail dependence structure among them.
Nicola Gnecco is a UC Berkeley postdoctoral researcher in the Bin Yu's Group. His research interests are causal inference, distribution generalization, and extreme value theory.
Previously, he did a postdoc at the University of Copenhagen, supervised by Jonas Peters and Niklas Pfister, and a PhD at the University of Geneva, supervised by Sebastian Engelke. He holds a master's in Statistics from ETH Zurich.
Causal inference
Abstract: This mini-course offers participants a dive into the foundations of causal inference and its recent advancements at the interface with extreme value theory. In the first class, we will introduce fundamental concepts in causal inference, such as directed graphs, structural equation models, interventions, Markov property, and faithfulness property. In the second class, we will first assume knowledge of the underlying graph and learn how to compute interventional distributions from observational ones. Then, we will study the problem of structure learning, where we try to identify and estimate the causal structure from the observational data. In the third class, we will study recent advancements at the interface of causal inference and extreme value theory for both iid and time series data.