Build separate relevance network plots for each Y group


I have built a relevance network plot from my DIABLO analysis, in which I integrated three data blocks against the output Y (0 = healthy, 1 = depressed).
I was wondering if it’s possible to build two separate relevance network plots for each group: one showing the inter-block relationships in the healthy group, and another showing the same for the depressed group. How would you achieve this?

Thank you in advance for your help.


hi @gc2416,
That would be a bit tricky to do, given that the variables are selected so as to discriminate your sample groups from the start, so it would be almost as if you were working backwards from there.

One manual solution would be to identify the variables that are (according to the sample mean or median) over expressed in a particular group, using the plotLoadings() function (which you can save as an object, i.e.

myobject <- plotLoadings(fill the arguments here)

Then identify the specific variables of interest, and then build a network using some other existing method. There might be an option to rerun a DIABLO (no variable selection) on this specific set of variables but that would be an overfitted model if you report the classification performance.