Estimation of the minimum detectable value for the determination of PCBs in fatty food samples by GC–ECD: a curvilinear calibration case

Electron capture detectors (ECDs) are known to have a limited linear calibration range. It was demonstrated that for ECDs, curvilinear calibration models yield more accurate results. To estimate the minimum detectable value (MDV), the ISO 11843-2 procedure was modified for curvilinear calibration graphs. At first, the graphs were linearized. The MDV was assessed with formulas based on the estimation of the MDV in the linear calibration case. Alternatively, the MDV was determined by a method based on confidence intervals for a second-order calibration plot. There was a close correspondence between the MDVs calculated by both methods. Weighted regression models were used to correct heteroscedasticity. It was found that applying weighted regression decreased the MDV with a factor of ≥10. It was concluded that the presented modifications to ISO 11843-2 are valid alternatives to estimate the MDV when dealing with curvilinear calibration graphs, but that heteroscedasticity has a greater impact on the estimated MDV than the non-linearity. Therefore, weighted regression must be used when heteroscedasticity is suspected.