sPLS choice of optimal number of components


I have been mixOmics and I am really impressed with the analysis I can do. I am using sPLS and I have a question. I run perf first to select the optimal number of components using PLS. Then I run sPLS using the ncomp calculate from perf function and I also tuned the optimal number of keepX variables. After I run tune.spls I got outputs tune.spls.MAE $choice.keepX for optimal number of keepX
When I build the final sPLS model shall I use ncompspls or the ncomp from perf function?

Many thanks,

Great to have you using our package! I would use as the number of final components the components indicated in ncompspls as this is your final sparse model (the number of component from the PLS tuning is a first indication only).
Note that we will be adding adding more functionalities on the tune function for s/PLS in the coming month with @aljabadi.


Thanks for your response!

Hi , do you have any update on this for PLS tuning component thanks.

Hi @amnah,

Please see Pipeline for (s)PLS: tune, model, perf? - #2 by aljabadi