Overview
Dani Gamerman is a Bayesian statistician at the Universidade Federal do Rio de Janeiro (UFRJ), Brazil. His research spans Bayesian computation, dynamic models, spatial statistics, and hierarchical models for non-Gaussian data. He is best known for the IWLS-proposal MCMC algorithm for GLMMs and for co-developing the theory of dynamic generalized linear models with H.S. Migon.
Key Contributions / Features
- IWLS-proposal MCMC for GLMMs (1997): Unified Metropolis-within-Gibbs scheme for Bayesian inference in generalized linear mixed models; WLS proposal N(m(t),C(t)) derived from IWLS linearization achieves 85–98% acceptance rates across binomial and Poisson applications with no tuning.
- Space-Varying Regression Model (with Moreira and Rue, 2003): Co-developed the multivariate pairwise difference prior for spatially heterogeneous coefficients; compared three MCMC schemes (single-site, block Normal-Gamma, analytical marginalisation, ranking C > B > A); applied to Amazon land use VAR (228 Brazilian counties).
- Dynamic generalized linear models (with Migon, 1993): Extended West-Harrison dynamic linear models to exponential-family responses; foundation for the DGLM framework used in Bayesian time-series analysis of non-Gaussian outcomes.
- Dynamic Bayesian survival models (1991): Extended the GLMM framework to survival/hazard data with time-varying covariates.
- Textbook on Markov Chain Monte Carlo (with Lopes): Markov Chain Monte Carlo: Stochastic Simulation for Bayesian Inference (2nd ed. 2006, Chapman & Hall/CRC) is a standard graduate reference.
Related