DoE augmentation

[#1]

I have a main effect screening design via custom design platform. Now I need to augment the design to model the curvature. There seems to be two options: (1) add axial points along with center points and (2) use Augment button and select RSM to add interaction and quadratic terms. 

Option 1 seems for central composite design. Can anyone tell me whether there is any particular reasons to use option 1 or option 2 above to model curvature in the model? Thanks.

[#2 · Accepted solution]

Hi,

Option 1 is the "classical" or "textbook" approach. This would be the common approach when computing and software were not available for designing (and augmenting) experiments.

Option 2 is the "optimal" or "custom" approach. This is the approach that JMP recommends as it enables you to design the right experiment to meet your objectives and solve your problems.

So the choice of option 1 vs option 2 is the choice of classical vs optimal approaches, which has been discussed many, many times over the last few decades.

Classical approaches are tried and tested. They are generally easier to understand.

Classical and optimal approaches will often converge on the same design (or augmented design) because classical designs are optimal.

Optimal approaches give you more flexibility. You can decide the number of runs. You could decide that you only wish to estimate quadratic terms for some of the factors. Or you could decide to only estimate some of the possible interactions. You can constrain the factor ranges and exclude certain regions of the factor space.

Option 1 will enable you to meet your objective to model the curvature. Option 2 would also work and give you more flexibility on how you get there.

I hope that helps,

Phil

[#3]

I recommend Optimal Design of Experiments: A Case Study Approach by Goos and Jones for anyone that wants to understand more about optimal design. The book includes case studies on RSM and augmentation. There is a free chapter download available here.