Choose between Plackett-Burman and Custom Design when having disallowed combinations

[#1]

Hello,

We want to run a screening DOE with 22 2-Levels factors, nearly all of them being categorical, and we wanted to use a Plackett-Burman with 24 runs in order to minimize the runs. But we have some combinations that are not possible to achieve, for example we will try a coating in either one or two layers and we want either to rinse or not between the two layers. But of course this rinsing makes no sense with only one layer.

So my question is: Can I generate a Plackett-Burman design and then change manually some combinations in the table to avoid impossible ones? And does jump take this change into account when analysing the data?

Or should I rather use a Custom Design where I saw that we can include "disallowed combinations"?

Sorry I'm quite new to DOEs so my comprehension of what is behind all kinds of models is limited.

Thank you very much

[#2]

Hi @PredictingSet44,

Welcome in the Community !
I would encourage you to use the Custom Design platform when dealing with disallowed combinations. Plackett-Burman is a classical "textbook" design very helpful but it can't handle any constraint.
You can create a Bayesian D-Optimal design with interaction effects estimability set to "If Possible" : https://www.jmp.com/support/help/en/18.2/#page/jmp/optimality-criteria.shtml#ww600960 

More infos on my previous LinkedIn posts:

https://www.linkedin.com/posts/victorguiller_doe-experimentaldesign-datascience-activity-72670847145...

https://www.linkedin.com/posts/victorguiller_doe-experimentaldesign-datascience-activity-72543860324...

Hope this answer will help you,

[#3]

Your situation sounds more like you need to consider a nested design.