shr264/pdag

simulation code for PDAG

PDAG

This contains all the simulation code for PDAG. The main function used in generating the tables is located in simulations.R. This requires installation of the partitionDAG package, which in turn requires the pdagDFS package. To install these packages, please type the following commands into R:

install.packages("devtools")
library(devtools)
install_github("shr264/pdagDFS")
install_github("shr264/partitionDAG")

Files

A few of the key files in this repo include:

• amat.Rdata
• data_generating_functions.R
• metric_functions.R
• simulations.R
• pDag_dairy_cattle_data.pdf
• pDag_dairy_cattle_data.Rmd

Details on key Files

amat.Rdata

This file contains the adjacency matrices for Yeast1, Yeast2, Yeast3, Ecoli1, Ecoli2 from the DREAM3 challenge.

data_generating_functions.R

This file contains the files to generate the true covariance matrix according the various adjacency matrices incluiding those from the DREAM3 challenge as well as random DAGs.

metric_functions.R

This file contains functions that calculate various metrics of interest, especially the macro-averaged AUC for DAGs.

simulations.R

This file contains the code to generate the simulations reported in the paper.

pDag_dairy_cattle_data.Rmd

This file contains the code to conduct the analysis of the dairy cattle data.

pDag_dairy_cattle_data.pdf

This file contains the output for the analysis of the dairy cattle data.

Example

To generate all the data for the tables, open up bash and type:

Rscript simulations.R

In particular, the lines that generates Table 1 are:

# Table 1
values = expand.grid(
  list(
    Methods = c('pcalg_custom','ccdr_paper_t','partial2','pcalg_addBG2', 
                'partial2_t',
                'pcalg_addBG2_t','lingam_custom','pcalg_custom_par','pcalg_custom_stable'),
    nlambda = c(30),
    Btypes = c('genB_mult_Yeast1','genB_mult_Yeast2','genB_mult_Yeast3',
              'genB_mult_Ecoli1','genB_mult_Ecoli2'),
    Ns = c(40,50,100,200),
    Seeds = 1:10,
    m = c(2),
    m1 = c(25)), stringsAsFactors = FALSE)


table1 = mcmapply(get_metrics_by_method, method=values$Methods, nlambda=values$nlambda, Btype=values$Btypes, n=values$Ns, seed=values$Seeds, m=values$m, m1=values$m1,
                  mc.cores=ncores)

Python version

The Python version and related code is available at https://github.com/shr264/pyPDAG.

Authors

Syed Rahman