Response surface design with more than 8 parameters

[#1]

Hi fellow JMP users,

I’m trying to obtain a Response Surface model for a process. The process consists of about 24 factors and it’s known that the response of the process is nonlinear with respect to at least 10 of the factors. I have tried a screening design to eliminate some of the factors, but the best I could do is reducing it down to about 15. Any more reduction will likely lead to some of the key sensitivities being omitted. Since JMP only allows a maximum of 8 factors, I’m wondering if there are ways to get around this problem? Thank you very much!

[#2]

Hi,

8 Factors is the limit with the RSM classical design.  If you use the Custom Designer, you can create a RSM design with as many factors as you like.

[#3]

Hi HadleyMyers,

Thanks for the reply. I did notice the Custom Designer and the option to create RSM design with large number of factors. However, the exact method used for the design creation seems to be obscured by the user interface. For example, was it using Central Composite Method? Or Box-Behnken? I tried to create a RSM design with 17 factors and the default "number of runs" came out to be 177. That number seems very low. So I would like to understand how JMP was able to create such efficient design? Thanks!

[#4]

The principles of design of experiment are always the same, but there are many different design methods. The old Central Composite Design method and Box-Behnken Design methods were developed during World War II in the UK. They are still good designs, but they have been surpassed with newer methods,. JMP provides Custom Design, which is based on the coordinate exchange algorithm. See the documentation for more information.