Custom Design – problem with a categorical variable and 2 continuous variables

cbaril

Hello,

 

We want to optimize 3 factors:

• Compound X type: The type of compound (4 different types 1, 2, 3, 4)
• Phase 1 Compound X concentration: The concentration of compound X in phase 1 of the process
• Phase 2 Compound X concentration: The concentration of compound X in phase 2 of the process

 

The design must test 0 g/L concentrations.

 

When doing a custom design and defining Compound X type as categorical and Concentrations as continuous, we get this design:

 

 

Compound X type	Phase 1 Compound X concentration	Phase 2 Compound X concentration
2	22	0
1	11.55	6
1	22	0
2	22	6
4	0	6
4	22	6
2	0	6
1	0	3
2	0	0
2	11	3
3	0	0
1	22	6
1	0	0
3	0	6
4	0	0
3	22	0
4	11	0
3	22	6
4	22	3
4	11	6
1	0	6
4	22	0
4	0	3
3	11	3
2	22	0

 

There are 4 runs (marked in red) that correspond to the same condition (No compound at all in the process).  

How to tell JMP that these are the same conditions? Should this condition be performed only once and the same response value entered for these 4 runs? How to make the design taking into consideration that at 0g/L the Compound type does not need to be studied?

 

Thanks in advance for your input!

Claire

Dan_Obermiller

I would just delete the "extra" runs from the design. 

 

Another approach is to add a linear constraint that says Phase 1 Compound X + Phase 2 Compound X > 0. The constraint is actually for greater than or equal to, so you could put in a small value greater than 0 like Phase 1 + Phase 2 > 0.1. It would look something like this:

Note that you will still have runs where one of the concentrations will be 0, but you will not have a situation where they are both zero. So, you would need to add a run for the 0, 0 combination.

 

Either way, I think you will need to manually modify the design and that is okay. If you are concerned about how the alteration affects some of the design characteristics, make the modification and then evaluate the design from the DOE menu.