I have a CCD with 3 factors on 5 levels and 8 responses. I have 20 runs in total. In 3 out of 20 runs I cannot measure 2 of the 8 responses (they represent 1 factorial and 2 axial points).
My alpha level for factor A and B is 1.68. For factor C I chose alpha = 1, because otherwise I would exceed the range of operability.
I have 2 questions:
1) How do the missing values influence the statistical reliability of the model?
2) Is it possible to have different alpha values for different factors in CCD?