Models for Longitudinal Data with Applications to Early Warning Systems

This book is the result of the joint work of a group of researchers who participated in a project funded by the Italian Government (PRIN 2022—Projects of Significant National Interest) entitled “Hidden Markov Models for Early Warning Systems”.1 As the title suggests, the project focuses on forecasting rare events by relying on an important class of models for longitudinal data based on unit-specific sequences of latent variables following a Markov chain. Related models are also considered, together with measures of predictive performance, which are crucial to evaluate the proposed approaches. Applications span economics and health, as the proposed methods lie at the inter-section of Econometrics and Statistics and have a strong connection with Machine Learning (ML).