What inspired this wish list request?
If the profiler has categorical levels and the same set of levels do not exist across responses, JMP will not display the group profiler. It does not matter if those categories are from categorical factor(s) or blocks or even random effects that are not shown by default. We had to create a tool to split the group profiler to work around the issue because we run into this problem all the time. It wastes significant time because we have to hunt down which groups are missing for which responses and identify how to split up the profiler. Also, we can't perform simulations on the block of responses we want to, we have to run multiple simulations. For JMP users without coding skills, they have no choice but to perform multiple analyses to keep the response groups seperated.
What is the improvement you would like to see?
The improvement idea is simple, develop a way to show the profiler even when the categories are not consistent. With missing data, the same total set of categories are not changing, just which response has them according to missing data.
A frequent case is with mixed models where the random effect is not displayed in the profiler. Can the profiler at least be shown in the cases where the categorical factor with the missing categories is not shown? The display would not be impacted so it may be the easiest case to fix.
Why is this idea important?
The JMP profiler is arguably the most important tool JMP has. No matter what methods we use to construct models (inside or outside of JMP), our engineers want their hands on the profiler. It should not be limited by this somewhat common issue. I have to explain to our engineers multiple times a month that this is an annoying JMP limitation we have to work around.