## Code for the manuscript ``Partition MCMC for inference on acyclic digraphs''
#
## Jack Kuipers and Giusi Moffa
## ETH
#
## Last modified: April 19, 2015
#
## Disclaimer: The code in this archive is not guaranteed to be optimised or free of bugs.
##        Please report any issues to the authors 
##        (jack.kuipers@bsse.ethz.ch and giusi.moffa@gmail.com).

The R files in the main directory run the various MCMC schemes for simulated and real data as follows:

simpleexample.R runs a small example on three nodes where the score is proportional to the number of edges in the DAG

smallsimmain.R runs a simulated example with 5 nodes and the BGe score

The plots resulting from running the R files are stored in the corresponding directory with graphs appended after the identifier.

In the runningexamples directory, knitr code and the output pdf files for both these small examples are provided.

BHmain.R runs an example with the Boston Housing data with 14 nodes and the BGe score.

medsimmain.R, largesimmain.R and largersimmain.R run larger simulated examples with 14, 18 and 20 nodes respectively.

The underlying MCMC codes are stored in the directories edgerevandstructure and orderandpartition respectively with the initialisation and scoring codes in those directories.

To change the scoring for DAGs, the DAGcorescore function, for example in scoring/numedgescore.R can be modified to any function of the parent nodes.