I'm currently analyzing test-results and have problems fitting the correct curve over my Data-points. The exponential fit I'm using does not always seem to be the right one - see screenshots. The problem is that the Points are not evenly distributed along the x-axis and that also that the Points are sometimes more, sometimes less scattered. What exponential fit would you suggest to use? The aim is to see at what point a (more or less) stable level is reached.
I'm currently analyzing test-results and have problems fitting the correct curve over my Data-points. The exponential fit I'm using does not always seem to be the right one - see screenshots. The problem is that the Points are not evenly distributed along the x-axis and that also that the Points are sometimes more, sometimes less scattered. What exponential fit would you suggest to use? The aim is to see at what point a (more or less) stable level is reached.
Hi,
The distribution of points of x-axis coudn't affect on fitting result. In this case wrong fitting function or/and parameters would give you bad result. Could tell wich function did you use, did you change parameters? Exponential funtion like "expdec1" from basic origin function in your case must be fine.
1) As Nikolay suggested, you can choose the "best" exponential decay model function. How about ExpDecay2/3 (or ExpDec2/3)?
2) If changing model is not desirable theoretically, try "Orthogonal Distance Regression" at the Iteration Algorithm option, which eases the effects of the y error at the almost-vertical area. (Though this requires Pro version.)
finding best fitting method
Posted - 08/26/2017 : 10:44:48 AM
