Hello,
I need your support for the following question.
For a sake of time, we have performed a central composite design (2 factors, 5 levels, 2 responses one to maximize and one to keep lower than an upper limit, 40 runs) using a bench-top model (before qualification) to identify our factors ranges.
In the meanwhile, a qualified model was ready to perform additional runs, so, we performed confirmatory runs for those proposed ranges (9-10 runs) using the qualified instead of the bench-top model. We identified an offset between the 2 models.
For those 9-10 runs that we have with the qualified SDM, could we put them into their own model (since it is a qualified model, while the other bench-top model is not and differs in some ways). However, rather than doing runs in the center of each range, the “set point runs” will be skewed to the lower left-hand region of the factor ranges. Can JMP accommodate this and give us (for example) predictions for failure rates at each corner of the ranges using the qualified model ?
Thanks a lot for your support!
Best,
I
The original DOE (central composite design) supported fitting a complex model (linear predictor) with first-order, second-order, and interaction effects. (Note that there are other design methods that result in other designs that would also support this model.) Using only the confirmatory to fit the same model might pose regression problems. The confirmatory runs are not necessarily sufficient to support fitting such a model well or at all.
Can you share the two data sets as JMP tables? Can make the data anonymous?
