When one of the rows of the proposed DOE data table gives conditions that cannot possibly be run, what are my options? Can the data be analyzed without that condition? Can I replace that particular row with some other set of conditions that could be run?
It may be possible to analyse the results without the row, depending on the number of runs and model structure, but a better approach would be to design the experiment taking account of the constraint that you are experiencing. If you use the custom designer then you can specify one or more constraints that will prevent the design generating runs that cannot possibly be run. Once the constraints are defined, the algorithm will generate an optimal design subject to the constrained design space.
ps: the link below is for a video that might be of use (I've not watched it myself, but from the title it should be relevant - and who better to learn about the custom designer than Brad) :-
Thanks for this. Indeed, I was able to build constraints based on two-factor interactions readily, however, the three or more factor interactions are harder to build constraints around given the linear nature of the constraint formulae. I have 12 constraints built in so far, but unfortunately am still coming up with one or two runs that are not possible. But, as mentioned, I will increase the number of runs and hopefully that will allow me to analyze the result. I almost need to perform DOE to build the constraint equations...