Practice JMP using these webinar videos and resources. We hold live Mastering JMP Zoom webinars with Q&A most Fridays at 2 pm US Eastern Time. See the list and register. Local-language live Zoom webinars occur in the UK, Western Europe and Asia. See your country jmp.com/mastering site.
See how to:
Understand importance of Multiple Response Optimization, Trade Space Analysis
Answer important questions as design prerequisites
Examine first example - four continuous factors, three responses, and 2nd order RSM model
Build second example - continuous, discrete numeric, categorical, and hard-to-change factors, added constraints, and 2nd order RSM model
Questions from the session answered by Tom Donnelly @tom_donnelly and Phil Kay @Phil_Kay
Q: What is the actionable solution to the pizza design?
A: The next step is to run the experiment. I did not do that. However, here are the instructions that a real pizza manufaturer gave to its pizza bakers after running the experiment.
Q: For the first example, the color zones were just added using the axis settings?
A: Yes. You can add spec limits to your response variables (https://www.jmp.com/support/help/en/16.1/#page/jmp/spec-limits-column-property.shtml) and have them appear on your profiler plot, but you would need to edit the axis settings (https://www.jmp.com/support/help/en/16.1/#page/jmp/customize-axes-and-axis-labels-in-graphs.shtml#ww...) to show a colored spec range.
Q: What are the capabilities of JMP to design clinical trials? I see that JMP can analyze data from clinical trials, but I am not sure if we can design them.
A: JMP currently is not used much for design of clinical trials. However, there is no reason why it could not be used. Clinical trials are different from industrial R&D experiments. In some respects, they are simpler because there is often only one factor: drug or placebo. The more complex aspects of clinical trial design might be something you for which JMP could be usefully used.
Q: How did you set constraints?
A; There are a few options. I used Uncoded constraints (raw units). another options is to set all your factors on a range of minus one to one then use coded units. Or you could ask JMP, for example and put in the two points that define these two lines (the upper and lower constraints). Then, in this example, JMP will solve, and you proceed from there to specify interactively. For advanced users, you can also apply domain expertise and put in equations for circles and ellipses, and you can get very exotic on your constraints, you get a more than one constraint, etc. There are also ways to set up inequality constraints that indicate unacceptable values (like never use this type of meat for a low-calorie pizza).
Q: What are whole plots and when are they used?
A: The point of whole plots is to handle factors that are hard to change, like in the pizza study, temperature. You don't want to change that factor between every single run. So, within the whole plot, you will have a series of runs where temperature is maintained as a constant and you will only change temperature between whole plots. You have the option to decide how many of these whole plots you want, based on how many times you can and then want to change that hard-to-change. Like many things in design of experiments, it's a balance and a compromise.
Q: What is the best design to start with?
A: Custom Design is a good, flexible choice and lets you do screening designs, but it turns out that there are certain types of screening designs that can be more efficient, like the super saturated designs that were introduced in JMP 16. Or you may use Definitive Screening Designs, but it may turn out as soon as you have you know categorical factors, with more than two levels, the definitive screening is out. (See attached Donnelly Summary of JMP Design Options Uses and Benefits.pdf.)
Q: What are the minimum and maximum number of factors to develop the perfect model?
A: There is no fixed number, of course. In the beginning stages, you often have many factors, and you must consider several questions that I outline in the blog. Making assumptions without addressing the assumptions can potentially leave the solution on the whiteboard. If you can, it's better at the beginning to really look at lots of factors and screen them and that's it and that's efficient experimentation. JMP also lets you do supersaturated and screening designs. See the attached slides for other JMP DOE Platforms.
Resources