cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Check out the JMP® Marketplace featured Capability Explorer add-in
Choose Language Hide Translation Bar
YanivD
Level III

Mixture design doe

I want to do a mixture design for a formulation with 4 components. I have two suppliers for the main component. Should I do one DOE with one categorical variable, or a separate DOE for each supplier?"
2 REPLIES 2
Victor_G
Super User

Re: Mixture design doe

Hi @YanivD,

There might be no definitive answer to your question, as both approaches may have benefits and drawbacks. It greatly depends on your goal, the complexity of the model assumed and the experimental budget you have.
As an example, if you want to compare directly and easily these two suppliers, the option to do 2 DoEs is interesting, as you will be able to compare raw values in the experimental space as well as predictions and models parameters. But depending on the mixture model's complexity, this could represent a high experimental cost.
On the other hand, if you want to be able to optimize your mixture depending on the supplier, creating a mixture design with the factor "supplier" might be helpful, as you can take into consideration in the model the interactions between supplier and other mixture factors.

If you want more guidance, you can describe a little more what is your objective, model, constraints... as it may help us figure out which way might be interesting to consider.
I hope this first answer will help you,

Victor GUILLER
L'Oréal Data & Analytics

"It is not unusual for a well-designed experiment to analyze itself" (Box, Hunter and Hunter)
statman
Super User

Re: Mixture design doe

Victor has given great advice and more importantly some questions.  Typically, mixture designs are optimization strategies.  These should be used after you understand noise (e.g., consistency of raw materials, supplier-to-supplier variation, etc.).  If you do not understand the noise, your "surface" created by the mixture design which is used to select appropriate levels for the mixture factors may change over changing noise.  This renders the model useless as the noise will invariably change in the future.

"All models are wrong, some are useful" G.E.P. Box