Using INLA to fit a complex point process model with temporally varying effects - a case study
Integrated nested Laplace approximation (INLA) provides a fast and yet quite exact approach to fitting complex latent Gaussian models which comprise many statistical models in a Bayesian context, including log Gaussian Cox processes. This paper discusses how a joint log Gaussian Cox process model ma...
Main Authors: | , , , |
---|---|
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
UCLA Statistics
2012
|
Subjects: | |
Online Access: | http://eprints.gla.ac.uk/199441/ http://eprints.gla.ac.uk/199441/1/199441.pdf |
Summary: | Integrated nested Laplace approximation (INLA) provides a fast and yet quite exact approach to fitting complex latent Gaussian models which comprise many statistical models in a Bayesian context, including log Gaussian Cox processes. This paper discusses how a joint log Gaussian Cox process model may be fitted to independent replicated point patterns. We illustrate the approach by fitting a model to data on the locations of muskoxen (Ovibos moschatus) herds in Zackenberg valley, Northeast Greenland and by detailing how this model is specified within the R-interface R-INLA. The paper strongly focuses on practical problems involved in the modelling process, including issues of spatial scale, edge effects and prior choices, and finishes with a discussion on models with varying boundary conditions. |
---|