My actual dataset is much more conplicated, there are many other columns beside the ones I listed, for more detail in the "by" group, let me describe as below: PARM x y Field ParmName F_ParmName Filter 3 x0 y0 1A AA 1A_AA pass 3.1 x1 y1 1A BB 1A_BB pass 2.9 x2 y2 1B AA 1B_AA pass 5 x3 y3 1B BB 1B_BB fail Field and ParmName is independent, I contact them as a new column F_ParmName, PARM is the actual value of each ParmName. The dataset is much larger so there are much more than one row of data for each category as I showed. I used by Field fitting, PaigeMiller's way is working but fitting result is wired since by group is not right, but when I try by F_ParmName fitting, the PARM '.' row will have prediction '.' result.
... View more