MCMChybridGP: Hybrid Markov Chain Monte Carlo using Gaussian Processes

Hybrid Markov chain Monte Carlo (MCMC) to simulate from a multimodal target distribution. A Gaussian process approximation makes this possible when derivatives are unknown. The Package serves to minimize the number of function evaluations in Bayesian calibration of computer models using parallel tempering. It allows replacement of the true target distribution in high temperature chains, or complete replacement of the target. Methods used are described in, "Efficient MCMC schemes for computationally expensive posterior distributions", Fielding et al. (2011) <doi:10.1198/TECH.2010.09195>. The research presented in this work was carried out as part of the Singapore-Delft Water Alliance Multi-Objective Multi-Reservoir Management research programme (R-264-001-272).

Version: 5.4
Depends: MASS
Published: 2020-11-12
Author: Mark J. Fielding
Maintainer: Mark J. Fielding <mark.fielding at gmx.com>
License: GPL-2
NeedsCompilation: yes
CRAN checks: MCMChybridGP results

Downloads:

Reference manual: MCMChybridGP.pdf
Package source: MCMChybridGP_5.4.tar.gz
Windows binaries: r-devel: MCMChybridGP_5.4.zip, r-release: MCMChybridGP_5.4.zip, r-oldrel: MCMChybridGP_5.4.zip
macOS binaries: r-release: MCMChybridGP_5.4.tgz, r-oldrel: MCMChybridGP_5.4.tgz
Old sources: MCMChybridGP archive

Linking:

Please use the canonical form https://CRAN.R-project.org/package=MCMChybridGP to link to this page.