GibbsPosterior-package: Samples from the Gibbs reference posterior distribution.

In josephmure/GibbsPosterior: Gibbs posterior

Description

In anisotropic Kriging models, estimating the correlation lengths is a difficult problem. The Gibbs reference posterior distribution is an Objective Bayesian solution. This package allows the user to sample from the Gibbs reference posterior distribution on correlation lengths.

Details

The GibbsPosteriorC functions samples from the Gibbs reference posterior distribution on the correlation lengths for Kriging models with Matern anisotropic (tensorized or geometric) kernel. It does this using a Metropolis-within-Gibbs algorithm.

Author(s)

Joseph MURE <[e-mail]>

References

J.Muré (2019), _Optimal compromise between incompatible conditional probability distributions, with application to Objective Bayesian Kriging_, _ESAIM: Probability and Statistics_, *23*, 271-309.

josephmure/GibbsPosterior documentation built on July 8, 2019, 11:33 a.m.