GPC calibration

Depending on the software, linear and/or polynomial regression models are supported. Strictly speaking, linear fits can only be used for linear columns. However, it is also observed here that a polynomial fit may be more suitable depending on the column performance. In addition, an odd-order fit is recommended.
However, there is no fit function that is equally suitable for all calibrations.

The most suitable fit function can be determined on the basis of 3 criteria:
1) The deviation between the calibration points and the calibration curve should be small.
2) The distribution of the residuals along the volume axis should be random.
3) The slope of the calibration curve should be physically meaningful, i.e. have no inflection points.

It is strongly recommended to try out different regression models to find the best calibration curve.