Abstract

This project focuses on exploring online changepoint detection, particularly concerning its application in identifying epidemic alternatives and detecting changes within large-dimensional models. We aim at proposing a range of techniques, referred to as "detectors," tailored for detecting changepoints in scenarios involving epidemic alternatives and large-dimensional models, as well as in cases where both phenomena occur simultaneously.

For the first objective, our aim is to develop a suite of techniques optimized for swiftly identifying the onset and conclusion of temporary changes. Regarding the second objective, our aim is to devise a collection of procedures adaptable to various regression setups, including those with large-dimensional data.

Our objective is to apply the proposed techniques across various fields of study. This entails promptly identifying epidemic alternatives through the analysis of Covid-19 hospitalization data. Additionally, we aim to validate our approach by examining data related to financial bubbles. We will also evaluate our methodology's capacity to detect breakdowns in pricing models using asset pricing data. Finally, predictive model breakdowns will be assessed using inflation prediction data.

Team

Professor Lorenzo Trapani – principal investigator 

Professor Carolina Castagnetti – co-investigator  

Professor Lajos Horvath – external investigator