| approximator-package {approximator} | R Documentation |
Bayesian approximation of computer models when fast approximations are available
Description
Implements the ideas of Kennedy and O'Hagan 2000 (see references).
Details
| Package: | approximator |
| Type: | Package |
| Version: | 1.0 |
| Date: | 2006-01-10 |
| License: | GPL |
This package implements the Bayesian approximation techniques discussed in Kennedy and O'Hagan 2000.
In its simplest form, it takes input from a “slow” but accurate code and a “fast” but inaccurate code, each run at different points in parameter space. The approximator package then uses both sets of model runs to infer what the slow code would produce at a given, untried point in parameter space.
The package includes functionality to work with a hierarchy of codes with increasing accuracy.
Author(s)
Robin K. S. Hankin
Maintainer: <hankin.robin@gmail.com>
References
R. K. S. Hankin 2005. “Introducing BACCO, an R bundle for Bayesian analysis of computer code output”, Journal of Statistical Software, 14(16)
M. C. Kennedy and A. O'Hagan 2000. “Predicting the output from a complex computer code when fast approximations are available” Biometrika, 87(1): pp1-13
Examples
data(toyapps)
mdash.fun(x=1:3, D1=D1.toy, subsets=subsets.toy, hpa=hpa.toy, z=z.toy, basis=basis.toy)