Subscribe Bookmark RSS Feed

Design of experiments help


Community Trekker


Mar 3, 2015

Fairly new to JMP here...

I'm trying to do a DoE with some complex systems.  For the sake of this question, let's say I have 8 recipes that I want to test for a specific response.  So a simple DoE would be to list each recipe as a factor (RecipeA, RecipeB, etc.).

But I would rather list each ingredient as a factor but only test the combinations of ingredients that lead to my recipes.  I'd also like to be able to play around with the amounts of each ingredient to see how that affects my response.

So let's say I have ingredients q,w,r,t,y

q,w,r = RecipeA

q,w,y = RecipeB

q,r,t = RecipeC

And each recipe produces a response.  But I wouldn't want to test r,t,y as that doesn't create one of the recipes I'm interested in.

How could I use JMP to do a DoE on this situation?  Could I design this experiment in JMP to also take into account the amounts of each ingredient?




Jun 23, 2011


I'm not sure I would handle it the way that you are suggesting. Instead I would probably run a Mixture Design on Recipe A with its three components Q,W,R. A second Mixture Design on Recipe B with its three components Q,W,Y and finally a third Mixture Design for Recipe C with its three components Q,R,T. This would allow you to leverage your domain knowledge with respect to acceptable ranges of the components for each recipe.