List.keepx for sPLS-DA

Hi, I am new to mixOmics.
I see,for sPLS-DA tuning parameter, we set up a grid of keep X values
Ex: list.keepX ← c(5:10, seq(20, 100, 10)). What are these values based on? Are we choosing them arbitrarily? What list.keepx I should set if I have 25 samples and 186 metabolites?

Thank you so much!

Hi @katea909,
We do not really have rules on how to set the grid. I would say that choosing the appropriate grid is a balance between the research question and waiting time (computational power).

Boundaries of the grid: When my research aim is to identify a minimal signature for prediction only, a small grid is obviously preferred (up to 10 or maybe 20). When I also want to interpret biology (e.g. by looking into protein-protein interactions), I use a larger grid (up to 50, 100 or even 150 in rare cases).

Fine vs coarse grid: This is a matter of how long you are willing to wait. Tuning the sPLS-DA is not really a demanding task, so I always use a very fine grid to get very precise results.

  • Christopher

Thank you, Christopher!
I have one more question regarding the PLS sample plot. The description says that if two blocks are not similar and the plotarrow has long lines, it indicates the lack of agreement between the two datasets. What exactly that mean? It means that PLS is not the best model in that case, or it means that there is no correlation between the two datasets?

Thank you,