Upcoming GRBIO seminars

Online. If you want to attend, please, contact grbio@grbio.eu

Improving dynamic predictions for multivariate longitudinal and time-to-event data via super learning and multivariate functional principal component analysis-based methods

Time-to-event and longitudinal data are common in health studies, and joint models (JMs) provide a way to analyze both while allowing dynamic predictions that are updated over time. Applying JMs to multivariate longitudinal data is computationally challenging. Ensemble methods like super learning (SL) combine algorithms to obtain optimal predictions and offer a solution for these problems. This work explores the use of SL for deriving dynamic predictions in multivariate JMs. Furthermore, alternative approaches for multivariate longitudinal data have emerged, notably multivariate functional principal components analysis (MFPCA), a flexible, non-parametric method well suited for longitudinal data. This study also explores the potential of combining MFPCA with machine learning methods to produce dynamic predictions.

Biosketch

Arnau Garcia-Fernández is an intern researcher at the Department of Statistics and Operations Research at the Universitat Politècnica de Catalunya - BarcelonaTech (UPC). Arnau graduated in Mathematics at the Universitat de Barcelona and is currently a master's student in the Master’s in Statistics and Operations Research (MESIO UPC-UB). Arnau is doing his master's degree final thesis directed by Prof. Dimitris Rizopoulos (Erasmus MC, Rotterdam) and Prof. Guadalupe Gómez Melis (UPC), and he will start his PhD at Erasmus MC this summer.