Special (Reduced) Scheffe Cubic Model for Mixture Designs

[Andreas_Büchse]

JMP offers the "Scheffe Cubic" Model as standard model for design and analysis of Mixtures. It is known from literature the these model tends to overfit the data. The "Reduced Scheffe Cubic" or "Special Scheffe Cubic" ofte is the better choice. It would very useful to have the Reduced Scheffe Cubic model as standard choice in the user dialogues dealing with mixtrure problems.

See: https://bookdown.org/gerhard_krennrich/doe_and_optimization/

See https://community.jmp.com/t5/Discovery-Summit-2015/Efficient-Mixture-Process-Experiments-Using-Nonli...

[Victor_G]

Hi @Andreas_Büchse,

I might be missing something, but does the "Simplex Centroïd Design" option in the menu "DoE", "Classical", "Mixture Design" would answer your needs ?
It seems to do exactly what is mentioned in your links.

You can also specify the model yourself if you're using the "Custom Design" platform.

I hope this answer will help you, 

[Victor_G]

Just a short comment to add an example of a Simplex Centroïd design case study I did: Use of Functional Data Explorer in a Mixture Design for Tribological Performance... - JMP User Commu...

Here is the design used (Simplex Centroïd design with K=3 and validation points to augment the design if necessary) : 

Hope these answers will help you build your DoE with JMP

[Andreas_Büchse]

Dear @Victor_G ,

thanks for your comment. You are right, that the simplex centroid in the DOE Design Platfrom leads to design which you can model with the reduced Scheffe-Cubic model.

My point for the wishlist is more related to the custom designer and the list of models to fit to data you have. Especially when you have additional process factors and need to fit mixture-process-models.

[Sarah-Sylvestre]

Hi @Andreas_Büchse, thank you for your suggestion! We have captured your request and will take it under consideration.

[mia_stephens]

Thank you for this suggestion. Goos & Jones outline two alternatives, and this would require some substantial research. We'll revisit as we begin planning for the next release.