Model calibration as a missing data problem: Approximate Bayesian Computation with MICE (Multivariate Imputation by Chained Equations)
MABC is an R package for calibrating stochastic simulation models to data. It implements a sequential Approximate Bayesian Computation method that employs MICE (Multivariate Imputation by Chained Equations) as an emulator to link model inputs to model outputs.
This repository holds the source code for the MABC package:
All code is written in R. R is a statistical programming language and software package that is distributed under a GNU General Public License. R documentation and software is available for free download through the R Project for Statistical Computing website at http://www.r-project.org. The software is available as an executable file for a wide variety of operating systems and computer architectures, and a compilable binary is also available should you need it.
MABC.R -- The core function to run MABC.
mice.parallel, model.mpi.run.R, model.parallelrun.R -- Helper functions used by the above two functions.
We have only tested this code on personal computers (OS X Version 10.11.6) and the golett cluster of the Flemish Supercomputer Centre (VSC).
Open MPI or Microsoft MPI Redistributable Package, and Rmpi. For installation instructions, see http://fisher.stats.uwo.ca/faculty/yu/Rmpi/.
R version >= 3.4.0
In addition, a long list of auxiliary R packages is required to run the functions in this package. All of these dependencies will get installed automatically when installing the software.
install.packages("devtools")
library(devtools)
install_github("wdelva/MABC")
All files are copyright protected and are made available under the GPL 3.0 License https://www.gnu.org/licenses/gpl-3.0.en.html. This means that this work is suitable for commercial use, that licensees can modify the work, that they must release the source alongside with Derivative Work, and that Derivative Work must be released under the same terms.
Please contact Prof Wim Delva with questions regarding the files provided and their authorized usage:
Wim Delva Email: wimdelva@gmail.com
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.