I'm trying to figure out why, when I click "Go" and look at the Step History, I see that JMP has chosen a 'best' model that actually has the highest, not lowest, AICc. I'm using JMP 12.1.0.
So, I tried with other response variables and this is not happening, but my response variable in this case was difference from the mean value, so the distribution was approximately normal and slightly off-centered from 0. When I run this response variable * 1000, the step history appears to behave normally.
