BRUG_BDMCMC

Overview

This pure C++ program is written to execute the algorithm in Bayesian Graphical Regression with Birth-Death Markov Process (BGR-BDMCMC) by Yuen et al. It is constructed and build upon the Gausian Graphical Model Double Metropolis Hastings Birth-Death Markov Process program(ggm_DMH_bdmcmc_ma).

BRUG_BDMCMC.cpp assumes the means are zeros for the Gaussian graphical models, whereas BRUG_BCMCMC_change_means.cpp further assumes a linear model for the means of Ys with respect to the covariates Xs.

Installation

BRUG_BDMCMC does not require any installation. It only requires the compiler g++ and the dependecies BLAS and LAPACK for C to be available.

Compilation and arguments

First, compile the code:

g++ BRUG_BDMCMC.cpp -llapack -lblas -std=c++11 -o BRUG_BDMCMC
# For the varying means version #
# g++ BRUG_BDMCMC_change_means.cpp -llapack -lblas -std=c++11 -o BRUG_BDMCMC_change_means

Next, run the compiled code:

# (Optional) For some devices, it may be necessary to update permissions.
chmod 777 ./BRUG_BDMCMC
# For the varying means version #
# chmod 777 ./BRUG_BDMCMC_change_means

# Create directory to store outputs.
mkdir demo_out
# For the varying means version #
# mkdir demo_out_change_means

# Run the program
./BRUG_BDMCMC -i 5000 -b 2500 -p 5 -L 4 -n demo_data/demo_n.txt -B 7 -g demo_data/demo_g_prior.txt -K demo_data/demo_K.txt \
-S demo_data/demo_S.txt -D demo_data/demo_D.txt -z demo_data/demo_beta0.txt -o demo_data/demo_beta1.txt \
-X demo_data/demo_X.txt -e .00000001 -s 123 -O demo_out/
# For the varying means version #
# ./BRUG_BDMCMC_change_means -i 5000 -b 2500 -p 5 -L 4 -n demo_data/demo_n.txt -B 7 -g demo_data/demo_g_prior.txt -K demo_data/demo_K.txt \
# -S demo_data/demo_S.txt -D demo_data/demo_D.txt -z demo_data/demo_beta0.txt -o demo_data/demo_beta1.txt \
# -X demo_data/demo_X.txt -Y demo_data/demo_Y_bar.txt -e .00000001 -s 123 -O demo_out_change_means/

The meaning of the arguments are as follows:

-i  Integer, the total number of iterations desired.
-b  Integer, the number of burn-in iterations desired.
-p  Integer, the dimension of the covariance matrix.
-L  Integer, the number of groups to be analyzed.
-n  String, the path to the file of the sample sizes by group.
-B  Integer, the initial value of b.
-g  String, the path to the file of g prior.
-K  String, the path to the file of initial K.
-S  String, the path to the file of S.
-D  String, the path to the file of initial D.
-z  String, the path to the file of \beta_0.
-o  String, the path to the file of \beta_1.
-X  String, the path to the file of X, the age of each group.
-e  Double, the value of the threshold.
-s  Integer, the value of the seed.
-O  String, the path to output. Please end it with a slash ("/").
# The following arguemtn Y_bar is only used by the varying means version.
-Y String, the path to the file of Y_bar. Y_bar is a L(row) x p(col) matrix of the by-group means.

Please note that:

• Files can be tab or space separated, but NOT comma.
• For non-symmetric arrays, the orientation of the files are important.
• Please kindly check the demo files for better information, or email me if there are any questions.

Demonstration

Please see the ./demo folder for the demo data

Testing

All testing is performed on CentOS Linux 7 and CPU Intel(R) Xeon(R) Gold 6130 CPU @ 2.10GHz

Inquiry

Please kindly email [email protected] for any inquiry or bug report, thank you!

Reference

Mohammadi R, Wit EC (2019). “BDgraph: An R Package for Bayesian Structure Learning in Graphical Models.” Journal of Statistical Software, 89(3), 1–30. doi: 10.18637/jss.v089.i03.