Preparation of calibration curves

[bluelotus]

Hello!
I have a question about how to construct my calibration curves.
My vinegar sample went through a procedure of ion-pair extraction cleanup before injection for amine analysis. There can be two choices of preparing the calibrators:
(1) standard solutions of amines (prepared in water) are analyzed directly as calibrators;
(2) standards solutions are first subjected to the ion-pair extraction procedure then analyzed.
Both of those practices can be found in the literature but which is correct (or better)?
I suppose the ion-pair extraction procedure conducted in (2) would bring about some uncertainties in the x-value (i.e. the quantity of analyte to be injected). Would this violate the assumptions of the method of least squares?
Thanks!

[tom jupille]

As a general rule, I try to match my calibrators and samples as closely as possible. That minimizes the potential for bias from matrix effects or sample prep, at the cost of increasing random error (as you point out).

As a minimum, I would compare the two (i.e., determine the recovery of the "prepped" standards). It it's essentially 100%, then you can go ahead and use "unprepped" standards.

[bluelotus]

Thank you Tom!
My spiked recovery studies show that the recoveries of the extraction procedure varied between ca 50% and 95%, depending on analyte. Is it OK to make calibrations as in (1) and then to use the recovery information to correct analytical results?

[tom jupille]

If I were a reviewer, I would accept that only if:
1. You could show that a significant (unsuccessful) effort had been made to improve the recoveries, and
2. You had data demonstrating that those recovery levels are consistent.

[Peter Apps]

Hello!

I suppose the ion-pair extraction procedure conducted in (2) would bring about some uncertainties in the x-value (i.e. the quantity of analyte to be injected). Would this violate the assumptions of the method of least squares?
Thanks!

What you want to know is the content of analyte in the sample, so the quantity of analyte to be injected is not important if you calibrate using standard that match the sample matrix and that are taken through the sample prep procedure.

Peter

[bluelotus]

Hello!

I suppose the ion-pair extraction procedure conducted in (2) would bring about some uncertainties in the x-value (i.e. the quantity of analyte to be injected). Would this violate the assumptions of the method of least squares?
Thanks!

What you want to know is the content of analyte in the sample, so the quantity of analyte to be injected is not important if you calibrate using standard that match the sample matrix and that are taken through the sample prep procedure.

Peter

Thank you for your reply peter.
Matrix effects were not supposed to take place, so I simply prepared the calibration standards in water.
My question is, if the calibration standards go through the whole sample prep procedure, both the x- and y-variables in the linear calibration model y=mx+b would be subjected to experimental errors,and, therefore, it seems questionable in statistics to calculate the regression equation using the least-squares method which minimizes the sum of the squares of the y-residuals only. Then is it necessary to use other algorithms to find the regression equation?

[Peter Apps]

The x values are the known concentrations of the analyte in the standard. They are no more variable than they would be if you simply made up solutions and injected them. The variability in the sample prep increases variability of the peak areas (y), not in the known content of analyte in the standard.

Although "Matrix effects were not supposed to take place" amines in vinegar are surely an analyte - matrix combination where they would be expected - this is probably why you get such different recoveries for different analytes. At the least I would make up my standards in dilute acetic acid.

Peter