Hi JMP community,
I have a 48-run DOE with 3 continuous factors, each with 3 levels, and 1 categorical factor, call it material, with 2 levels. I am interested in maximizing my response, let's call it yield, for both material 1 and material 2. Not only do I have strong reason to believe that the 2 materials will produce different yields, but that their response surfaces will be different. Therefore, I design for a model with all response surface terms and their interactions with material; 20 terms with the intercept.
My goal is to simultaneously find optimal settings for both materials and decide which material to use. The most obvious thing to do is to lock material at each level and maximize desirability, thus producing optimal settings for both materials. I can then compare these predictions.
I am looking for guidance on how to more formally test the difference in the materials, ideally with confidence intervals. Would I use Custom Test for this? Should I compare the bagged predictions at the 2 optimums?
An example of one of the questions I am trying to answer with this is "is material 1 15% better than material 2?" My idea was to estimate a confidence interval on the difference in predicted optimums and see if the interval contained my acceptance criteria. I realize that there are a few different ways to go about this problem; comparing predicted max yield or comparing average yield over a specific range of my factors. Would appreciate any other thoughts you all have on this problem.
-Robert
