A0152
Title: Joint modeling of complex data with latent variables
Authors: Xinyuan Song - Chinese University of Hong Kong (Hong Kong) [presenting]
Abstract: Several joint modeling approaches are introduced for analyzing complex data with latent variables. The models under consideration include a latent factor-on-image regression model to investigate the associations between imaging predictors and a latent outcome, a mediation analysis model to examine the causal effects of interest in the presence of latent mediators, and various survival models to reveal observed and latent risk factors for time-to-event outcomes. Statistical methods, including functional principal component analysis, the expectation-maximization algorithm, estimating equation method, spike-and-slab procedure, and Markov chain Monte Carlo sampling techniques, are used to conduct statistical inference. Applications to real-life studies regarding Alzheimer's disease and the complications of type 2 diabetes are presented.
