Name: Iván Esteban Gutiérrez
Institution: Pontificia Universidad Católica de Chile, Chile
E-mail: gustavo.rocha@ibge.gov.br
Co-authors: Danilo Alvares
Abstract:
Linear mixed-effects models are widely applied in longitudinal analyzes, where the latent structure (also known as random effects) has the role of hierarchically connecting the individual repeated measurements. In ongoing studies, this class of models requires a dynamic approach to quickly update the inferential process. Hence, we propose a sequential Monte Carlo strategy adapted to update the posterior distribution of fixed effects based on the marginal likelihood of mixed effects models and then get the random effects from their full conditional posterior distribution. From simulation studies, we have noticed that by gradually including new observations, our methodology turns out to be significantly faster than cutting-edge alternatives (for example, Stan).