Overview
Håvard Rue is a Bayesian statistician at the Norwegian University of Science and Technology (NTNU), Norway. He is best known for fast computational methods for Gaussian Markov random fields (GMRFs) and for developing the Integrated Nested Laplace Approximation (INLA) framework for approximate Bayesian inference in latent Gaussian models.
Key Contributions / Features
- Fast sampling of Gaussian Markov random fields (2001, JRSS-B): Exploits the sparse band structure of the GMRF precision matrix via Cholesky decomposition; enables block-update MCMC for spatial models with thousands of locations at manageable computational cost. Used by Gamerman-Moreira-Rue (2003) Scheme B.
- Space-Varying Regression Model (with Gamerman and Moreira, 2003): Co-developed the SVRM pairwise difference prior framework; contributed the sparse matrix computation for the block scheme.
- **Block updating in MRF models for disease mapping (with Knorr-Held, 2002, Scand. J. Statist.)**: Developed efficient block-updating strategies for GMRFs in disease-mapping applications.
- Integrated Nested Laplace Approximation (INLA, with Martino and Chopin, 2009, JRSS-B): Deterministic approximate Bayesian inference for latent Gaussian models; bypasses MCMC by combining Laplace approximations with sparse numerical linear algebra; has become the dominant method for spatial statistics and geostatistics applications where MCMC mixing is slow.
Related