Hey guys, I'm new here and somewhat new to JMP so please bear with me. I am looking to set up a DOE and I’m having a few issues. I am looking to modify an existing experiment conditions (concentration conditions) and have that ratio condition target a "known or standard." I have run multiple POC experiments and dialed in on a specific recipes that I would like to tweak or fine tune by exploring different reagent concentrations while still focusing on replicating results from the standard. I hope this makes sense, can anyone point me in a direction or have any suggestions?
Your description is not specific or clear enough for me to be sure of what you are asking so my reply might be far from the solution but I will try nonetheless.
I assume that you used JMP to design your experiment. I assume that you used custom design. I interpret your question as: how do I include a control run? You can manually add runs or manually change factor levels in the data table after you design of the experiment.
If you want JMP to add runs, select Augment Design. The new runs may use different factor ranges and new model terms.
One caution: when I hear words and phrases like 'dialed in,' 'specific recipes,' and 'tweak or fine tune,' I worry that you are thinking of a test and not an experiment. The purpose of a test is to get an answer. The purpose of an experiment is to fit a model in order to get lots of answers. Both approaches are useful and valid but they are different. I would never narrow factor ranges in an experiment, even if I had a good idea about where the best conditions occur.
Yes, the experiment was designed with the DOE function. To try and be clearer, I will be very specific with the situation. There was a control that was experimentally ran along with the multiple conditions. We are talking about changing concentration ratios. The results from one condition is very close to the standard, when plotting a correlation slope however, there is latitude between the ratios that I want to explore. I very well could be confusing test with experiment. I hope this somewhat helps.
I'll try to add a bit of a different perspective from my colleague @markbailey's advice...which is good advice to begin with. I'm focusing on your comment, '...run multiple POC experiments...'. One of my former stats instructors admonished us that BEFORE you experiment in the lab, see if someone else has plowed that ground before. Well if you've already run 'multiple' experiments, can you solve your practical problem by modeling/analyzing those results? It sounds like these 'experiments' might just really be a collection of single trials, without the unique, intentional, planned structure of a designed experiment, so if that's the case modeling might be more problematic, but it's at least worth some investigation perhaps? Then perhaps all your subsequent 'experimentation' is really along the lines of verification/confirmation and not what I'll call 'new knowledge building' that would come from a more formal planned DOE centric problem solving effort.