Speaker: Denis Rustand, Bordeaux Population Health Research Center
Abstract: Modern biomedical research data is increasingly complex, requiring advanced models for longitudinal outcomes, time-to-event data, and their joint analysis. While Bayesian methods are well-suited for these, traditional MCMC estimation is often computationally demanding, limiting the complexity of models that can be applied in practice.
This presentation introduces a unified framework that relies on the Integrated Nested Laplace Approximation (INLA) method to address these computational challenges. INLA offers a significant computation speed advantage over MCMC while maintaining high accuracy. The framework is applied to a comprehensive range of models, from advanced survival (e.g., multi-state, cure fraction) and complex longitudinal (e.g., zero-inflated, ordinal) models to their multivariate joint modeling.
The R package INLAjoint is presented as a user-friendly interface for this framework. We demonstrate its computational efficiency, which facilitates key applications such as the dynamic prediction of event risk based on evolving longitudinal outcomes. These applications will be illustrated, highlighting the INLA-based framework as a valuable and robust tool for analyzing complex health data.
Speaker bio: Dr. Denis Rustand is a Research Scientist in Biostatistics at the Bordeaux Population Health Research Center (Inserm U1219), France. He earned his Ph.D. in Public Health and Biostatistics from the University of Bordeaux, focusing on the development of joint models in oncology. Following his Ph.D., he was a postdoctoral research fellow working with the developer of INLA, Professor Håvard Rue. He is the developer and maintainer of the INLAjoint R package. His research focuses on Bayesian computational statistics, survival analysis, joint modeling of complex health data, and their application to medical research.
This will be a free online seminar. To register to attend virtually, please click here: https://cam-ac-uk.zoom.us/meeting/register/ab-78rraTl6lgongAI4C4g