GS/MMPC algorithm removed all non-related variable about causaldiscoverytoolbox HOT 4 CLOSED

Hi,
Don't be sorry! This bug was unexpected; it should be fixed in 0.5.17 !
Best,
Diviyan

from causaldiscoverytoolbox.

Hello @Cby19961020 and thanks you for noticing this point. This behavior is not what we want: even if a variable is uncorrelated, we still want it to appear in the adjacency matrix !
I will look into it this evening.
Best regards,
Diviyan

from causaldiscoverytoolbox.

After checking, it doesn't seem to be the x: nx.adjacency_matrix(x).todense() that is causing the issue, but from the algorithms themselves ?

from causaldiscoverytoolbox.

Hi Diviyan @Diviyan-Kalainathan ,

Thank you very much for your promot respond! I think you are right, maybe it is the algorithms that I am using.

Essentially the data I am working with is very similar to the "sachs" data used in the tutorial. Here is a screenshot. I have 54 input variables.

When I implement algorithms like GS or MMPC to my dataset I will only end up with 32 variables. These variables all have some sort of connection with other variables, as indiciated in the graph. The onces that are completely independent are ignored.

The nx.adjacency_matrix(output).todense() function will generate a matrix with no label, in this case a 32 by 32 matrix is generated. However based on just this I cannot figure out which one is variable Task_0 and which one is Task_1 etc.

Sorry for not stating the problem clearly and thank you for your hard work!

Best Regards,
Bo

from causaldiscoverytoolbox.