Dear JMP & Community,
I'm running JMP Pro 16.0 on a Windows 10 machine, 64-bit.
I've recently come across a strange potential bug in the Sort function under Tables. When I try to sort the attached data table using the three columns Training R2, Validation R2 and Diff R2, I don't get the output that I would expect.
If I sort it:
Big to small Training R2, then big to small validation R^2, then small to big Diff R2, I get the following (just a screenshot):
You can clearly see that the validation R2 is actually getting bigger and the R2 Diff is getting smaller. It's as if it only sorted by Training R2 and forgot the rest.
What I would expect is something closer to this (when I sort ONLY small to big R2 Diff):
Even then, it's not quite what I want. What I'm really looking for is having JMP sort out the data so that it first arranges the rows based on the hierarchy that Training R2 is big, followed by Validation R2 big, followed by R2 Diff being small. It used to work this way in previous versions of JMP, but now it seems to be different.
What I'm really hoping to get from the Sort function is for it to arrange the data table such that it finds the entry with the largest Training R2 AND largest Validation R^2 AND smallest R2 Diff. The normal way of doing this by casting the columns as such (below) isn't working like how it did or I think it should.
Our organization hasn't upgraded to 16.1 yet, so maybe this has been fixed. I only have 100 rows in this data table, but sometimes I have several thousand to sort through.
Any feedback much appreciated.
Thanks!,
DS
I just tried sorting it in 16.0 and 15.2 and got the same results in both places. When you sort by multiple columns it sorts by the first column first and then, within ties of that first column it sorts by the second column. So, you shouldn't expect the second column to be strictly descending.
I believe the Validation and R2 Diff columns will only become important when needed to break a tie between rows that have the exact same Training R2. For example rows 2-3 have the same training R2 so JMP moves to validation R2 to see if there is a difference there and so on. This is the behavior I would expect.
