We are currently applying for a research grant and today we found different JMP versions (i.e. JMP 7 vs JMP 10) give different R2 and final p-values using Fit Model where one of variable was set as a random effect. Why is this happening and is there way (such as turning on/off option somewhere in the preference menu) to unify the results?
I can't tell you why, for certain, but I would expect it has to do with the REML functionality (if you're using REML) being improved in later version. Additionally, there may have been some parameterization changes in later version. I know for the same data analyzed in JMP and another package, I can get different VIFs, different R^2, etc... none too greatly different, but sufficient that it causes me to scratch my head.