DOE Custom Design Conituous Factor Quadratic

[#1]

I have a DOE that I have setup to test for curvature in a continuous factor and the centerpoint that JMP selected is different than what I want to test.  For example:

Low: 20

Mid (JMP): 60; 50 is what I want to change it to

High: 100

I can manually adjust the values in the Table of the designed study but I am concerned that the Design Evaluation would then not be accurate (i.e. Power Analysis, Fraction of Design Space, VIF)

Is there a way to adjust the design earlier so that my Evaluation is accurate?

[#2]

You are correct that the Design Evaluation is not accurate. So, create the Design Data Table and make your changes. Then go to DOE > Design Diagnostics > Evaluate Design to get the correct evaluation associated with your design.

Alternatively, you could specify your factor to be a Discrete Numeric where you specify the level. Make sure to change the quadratic to be Necessary instead of If Possible. 

[#3]

Thanks Dan.  That works well...to make adjustments and Re-evaluate design.

My only issue now is that I cannot make a table with simulated responses and evaluate VIF on Fit Model/Parameter estimates.  I like to see that VIF is < 3 in addition to Design Evaluation Power Analysis and FDS.  However, I think this will still work quite well.

[#4 · Accepted solution]

If you want to check VIFs, just fill in your response column with any value. They can all be the same value even. Then ignore everything except the VIF numbers. The VIFs do not depend on the response values.

Keep in mind that the correlation of estimates is designed to give you an idea of the VIFs.

[#5]

Thanks so much

[#6]

I'm a bit late to the party, but I have a question:  Why are you purposely setting a Center Point (which is intended to be the center of ALL continuous variables in the study) off-center?  This introduces a bias that must be accounted for in the analysis.  If the design is orthogonal, your VIF's will of course be 1.

[#7]

The offset center point is to include the current setpoint in the study design.

[#8]

Then you should change the other levels so the design is balanced.