Introducing BACCO, an R Bundle for Bayesian Analysis of Computer Code Output

Abstract

This paper introduces the BACCO bundle of R routines for carrying out Bayesian analysis of computer code output. The bundle comprises packages emulator and calibrator, computerized implementations of the ideas of Oakley and O'Hagan (2002) and Kennedy and O'Hagan (2001a) respectively. The bundle is self-contained and fully documented R code, and includes a toy dataset that furnishes a working example of the functions. Package emulator carries out Bayesian emulation of computer code output; package calibrator allows the incorporation of observational data into model calibration using Bayesian techniques. The package is then applied to a dataset taken from climate science.