Skip to main content
Michael Ellis
← Back to open source projects

R · Open source

gaussian-process

Full Bayesian Gaussian Process Spatial Models

An R implementation of Gaussian process regression for spatial data using Gibbs sampling and adaptive Metropolis-Hastings.

Overview

This project fits hierarchical Gaussian process models to spatial data. Regression effects, process variance, measurement error, and the spatial range are represented separately in the model.

The sampler uses conjugate updates where they are available and adaptive Metropolis-Hastings for the remaining parameters. The example includes trace plots and credible intervals for a simulated spatial process.

Technical highlights

  • Exponential and Gaussian spatial correlation kernels
  • Adaptive Metropolis-Hastings tuning
  • Gibbs sampling for regression coefficients
  • Posterior spatial effects and credible intervals

Topics

Gaussian processesSpatial statisticsMarkov chain Monte CarloBayesian regressionR