PLS Modeling with mixture components

[#1]

I made a PLS on a 6 component mixture; as Design Roles I specified all components as mixture;

However, when maximizing desirability to find optimal composition, the profiler does not treat the  factors as mixture components cfr. below; 

How to switch the PLS factors to mixture components?

Will this also the case when I use ML predictive modeling? 

[#2 · Accepted solution]

Hi Franck,

See my answer on your other post: https://community.jmp.com/t5/Discussions/SVEM-ADD-IN-FOR-SUPPORT-VECTOR-MACHINE/m-p/919885/highlight...

Maybe the "Mixture" column property is missing for these mixture factors, so JMP doesn't know to which quantity/value these factors should sum up to ? This would explain why you can get mixture factors sum value > 1. 

[#3 · Accepted solution]

Can you verify that your mixture columns have a Mixture column property? The mixture constraint is reflected in the profiler when using PLS (and Bootstrap Forest, per your ML question) in the Donev Mixture Data sample data file, which has mixture constraints set up as column properties.

[#4]

Hi Jed & Victor, indeed this is missing, thanks a lot!