Paper Image

Bayesian modeling of coordinated dynamics in sparse longitudinal data

Paper Authors:

Tui Nolan,

Sylvia Richardson,

Hélène Ruffieux


Key Details

Presents first Bayesian framework for multivariate FPCA, with a shared score parametrization

Circumvents covariance estimation, enabling analysis of sparse/irregular data

Variational inference algorithm leverages fragment modularization for efficiency

Simulations show accuracy and computational advantages over alternatives

Analysis of COVID-19 data reveals coordinated biological dynamics linked to outcomes

AI generated summary

Bayesian modeling of coordinated dynamics in sparse longitudinal data

This paper presents a Bayesian hierarchical modeling approach for multivariate functional principal component analysis (mFPCA) of sparse longitudinal data. The model flexibly pools information across related variables and subjects through shared scores, enabling estimation of latent trajectories and dynamics. Fast variational inference is used for large problems. The method is applied to molecular markers in COVID-19, revealing coordinated immune and metabolic dynamics associated with disease severity and incomplete recovery.

Answers from this paper


No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up