Statistics Seminar - Bayesian Functional Regression for Extremes
| Dates: | 5 March 2025 |
|---|
| Times: | 14:00 - 15:00 |
|---|
| What is it: | Seminar |
|---|
| Organiser: | Department of Mathematics |
|---|
| Who is it for: | University staff, External researchers, Current University students |
|---|
| Speaker: | Dr. Nicolas Hernandez |
|---|
|
Dr. Nicolas Hernandez (QMUL)
Title: "Bayesian Functional Regression for Extremes".
Abstract:
Recent record-breaking extreme events-such as catastrophic wildfires, unprecedented flooding, and intense hurricanes-highlight the urgent need for improved statistical methodologies to analyse and predict such occurrences. Extreme Value Theory (EVT) provides a rigorous foundation for modelling extremes, yet conventional approaches often struggle with the increasing complexity of modern data, particularly when covariates take the form of curves or random functions.
In this work, we propose a Bayesian functional regression model for extremes, extending the Tail Index Regression framework of Wang and Tsai (2009) to accommodate functional covariates. Our approach builds upon ideas from Cardot et al. (2003), incorporating a Bayesian regularization scheme that balances flexibility and stability in estimating tail behaviour. The proposed model is evaluated through simulation studies demonstrating its practical effectiveness in extreme event prediction. This work represents an early-stage methodological development, and further investigations—both theoretical and empirical—are required to refine and assess its practical performance.
Speaker
Dr. Nicolas Hernandez
Organisation: QMUL
Travel and Contact Information
Find event
Frank Adams 2
Alan Turing Building
Manchester
