Fitting Geostatistical and Spatial Point Process Models to Spatial Survey Data

Data:
7 a 11 de Maio de 2018

Local:
Faculdade de Ciências da Universidade de Lisboa, Portugal

Formadores:
Janine Illian, University of St Andrews

Objectivos:
The course will cover methods for fitting geostatistical and spatial point process models to data obtained from surveys on which the whole region or population of interest is observed, as well as surveys on which observations are available only in a spatial sample from the region of interest, and surveys on which members of the population of interest are missed with some unknown probability. Inference for this last class of surveys will focus on distance sampling surveys. The workshop covers methods for spatial modelling problems in general, although the examples and exercises in the workshop will focus on ecological surveys.

Participants will be instructed in the use of the R package “inlabru”, developed by researchers at the Universities of St Andrews and Edinburgh, which is available here: http://www.inlabru.org/

This package provides a flexible and convenient interface for spatial inference with the R-INLA package, without having to be familiar with the details of R-INLA syntax and structures, as well as a means of doing spatial inference when detection probability is not known.

Participants will be instructed in the use of the software on example datasets, with some time being devoted to practical computer exercises in which participants use the software.

Participants should bring their own laptop to the workshop.

Participants should have some experience using the R statistical software package, and be familiar with basic statistical methods to the extent of being comfortable at least with linear regression models and preferably be familiar with fitting generalized linear models, and have a basic understanding of Bayesian inference.

The course will provide an introduction to the analysis of spatial data with the R statistical software.

• Spatial data types
• Bayesian Inference and spatial modelling
• Introduction to INLA and inlabru
• Prior specification and pc-priors
• Continuous domain random field models
• Finite element meshes for approximating continuous space
• Log Gaussian Cox processes
• Model choice and model validation
• Spatial sampling: thinned point processes
• Distance sampling as a thinned Poisson process
• Multiple likelihoods