Augmenting a DOE by adding one more continuous factor and removing two levels of a categorical parameter

Question

Ok, here's the DOE design that I have:

I have 3 continuous factors and 2 categorical factors

X1 (-1,1)

X2 (-1,1)

X3 (-1,1)

X4 (X4_1, X4_2, X4_3 and X4_4)

X5 (X5_1 and X5_2)

I would like to augment the design by adding an extra X6 continuous factor. And remove two levels of X4 and  one level of X5.

My DOE situation that I would like to have help with is as follows:

-It doesn't seem that I can remove levels in Augment design (eg, for X4, X5)

-I cannot add X6 in augment design (I can only manually create a new column and then in augment add another level)

To me, it seemed that creating a new custom design (as attached below: Follow-up design), gave me the lowest nb of extra experiments to run. I had to change what I put in the model to only account for X6, and the two extra questions to answer: X4*X6 interaction and X6^2

Is that the right approach? is Augmented design limited and I cannot do with it what I needed (aka, remove categorical levels and adding a continuous parameter)

Thank you

Yasmine

Follow up design-Florencia.jmp

Custom Design-Florencia.jmp

yasmine_hajar · Answer

I looked furthermore into this. There is actually a way around augmenting with only few levels of a categorical factor in augmented design.

The trick is to add a factor constraint and select the categorical levels that are not wanted

Using 'Or' instead of 'AND' for multiple constraints allowed to remove two factors from X4 and one factor from X5

My follow up question would be how to reduce the nb of extra runs added as I don't believe I need 8 experiments.

Augmented Design-Proper.jmp

yasmine_hajar · Answer

Thank you Jordan!

Jordan_Hiller · Answer

Hi Yasmine,

The reason it's not possible to augment as you describe has to do with properties of the data in your first experiment.

Deleting factor levels of X4 (or any factor) is not possible, only adding levels. Because you already have data for the levels you wish to ignore. What should JMP do when you launch fit model on the full augmented design? Omit the rows with the bad levels? That would throw out information about your other factors.

It's a similar issue with adding a continuous factor. If you were to start recording values for the new factor only after augmenting, the data would all be missing for the original rows, and Fit Model will throw them out. You can only add a new continuous factor if you have existing data for that factor already -- maybe it was held constant for the first study, or maybe it was uncontrolled but the values were recorded. In either case, you can augment by adding that existing data to the first design, then using it to augment.