Potential issues when computing the Markov blanket

[ivan-marroquin]

Thanks a lot for this great package!

When I computed the Markov blanket, I noticed issues with the p-values associated with the co-parent tests. Some of the p-values are higher than 0.05 (for a confidence of 95%), yet they are reported as False dependent tests.

I made simple additions to one of your original scripts to force reproducible results and display the found parents. In the attached zip file, you will find:

• the slightly modified version of the script to compute the Markov blanket
• my script to run the computation
• input file

Thanks for your help,

Ivan
test_files.zip