Publication at Faculty of Mathematics and Physics |
2022
Abstract
A regression model consisting of generalized linear mixed-effects models (GLMM) for diverse types of longitudinal outcomes. These seemingly independent structures are joined through the joint multivariate normal distribution for the vector of all random effects belonging to one subject.
Following the methodology of model-based clustering (MBC) the mixture of these models is created where the clusters differ in model parameters chosen in advance. Assuming special sparse finite mixture prior we face even the problem of unknown number of mixture components in advance.
The whole model including the classification of the subjects is estimated via MCMC methods which elegantly avoid troublesome integration of latent variables.