I have an experiment with 5 common variables (a combination of 2 level categorical and 2 level numerical). The 6th variable is glass or plastic, but they are not used together, it is either one or the other in each of the experimental runs. Here is the problem.
The glass has 2 numerical levels and the plastic is actually broken into two factors itself, a 2 level categorical and a 2 level numerical.
I thought about creating a dummy variable for the glass or plastic and then a second dummy variable for the glass levels and a third dummy variable for the plastic but I could not get it to work out. It's sort of like nesting a dummy variable within a dummy variable but the nested variables represent different factors. Nor could I figure out how JMP would analyze it anyway.
Any ideas would be appreciated.
Have you considered treating the Glass/Plastic factor as a 6 level categorical, namely G1, G2, P11, P12, P21, P22?
This would not allow for a DSD but you could use the custom design platform to perform this scenario.
I like Lou's approach because it is simple and direct. An alternative would be to use separate experiments, one for glass and another for plastic, since these choices are mutually exclusive, you would not be concerned about interactions on the choice of glass or plastic. These two experiments could be small if you are looking for large effects.
Currently I have 2 screening experiments set up like you suggest with 18 runs for each experiment. I was hoping to cut that number down with one larger one but less total runs. The runs are destructive and cost a lot of money.