Due to time constraints and experimental possibilities I am interested in instead of a completely randomized DOE in JMP, a 'shortest path' through my experimental space (but still be as random as possible):
1) minimize the number of cases where more than one parameter changes between two consecutive experiments
2) minimize multiple-step changes between levels within the parameters (i.e. prefer '-1 to 0' and '0 to +1' over '-1 to +1').
3) adding an extra experiment would cost less time than allowing either 1 or 2.
Has someone tried this / studied the effect of this on an experimental design and the statistical effects?
(maybe to visualize: considering visiting 10 cities by car to take a sample in each city in a 20-experiment DOE, minimizing the amount of petrol you want to use - not as the goal of the experiment, but as the design of the experiment).
(I am currently on JMP 13, but can get access to JMP 14 and 15).
Interesting concept. You'd definitely be playing in the Custom Design and Evaluate Design platform spaces. And probably do your work in JMP 14 or 15. As opposed to JMP 13 because the Evaluate Design platform, where you can efficiently compare multiple designs to each other, is enhanced in JMP 14 from 13, and maybe even more in JMP 15? I haven't seen JMP 15 yet so not sure.