Mixture with two factors - enough runs?

Question

Hello guys,

Im new in JMP and want to do a DOE for a school project. So I want to do a DOE for a mixture with two different compounds, so I have two factors. I did set the mixture ratio of X1 from 0.7 - 1 and X2 0 - 0.3 (X1 should have 100% as a reference).

But now I have for my experiment only two different mixtures. Will this be enough or is it possible to increase the number of different mixtures? Or should I limit the values of X1 to 0.7 - 0,9 and set the 100% reference experiment independently of this DOE? When I do this, this will give me three different mixtures as DOE.

Dan_Obermiller · Answer

The number of blends is based on the design principles.

You are using Custom Design. Custom design will choose the most efficient design TO FIT THE MODEL SPECIFIED. If you only specify a main effects model, that will be two levels as two points will define the line. In your screen capture you specified the interaction, which in mixture models is nonlinear blending (in other words, quadratic for this situation). Therefore, you get three levels. You could change the optimality criterion to be I-optimal which may  get you a few additional blends in the middle of the space, but in general, the design will be the fewest number of blends to effectively fit the specified model. Once the model is fit, the model is what is used to determine "optimal" blends rather than testing multiple levels to find the best.

If you truly want more blends for some reason, you might want to consider a space-filling design. JMP has several choices for that approach, too.

mlo1 · Answer

There were discussions on the need of mixture design .

May be that helps and another type of design is better.

https://community.jmp.com/t5/Discussions/DoE-Design/m-p/275541#M53466

https://community.jmp.com/t5/Discovery-Summit-2016/When-Not-to-Run-a-Mixture-Experiment/ta-p/24132