With the further explanation of how you generate different concentrations for your calibrations I really think that you are making this much more complicated than it needs to be.
Forget all about the uncertainty of the blank. The blank is outside your calibration range and so you will never be able to report a result for a blank.
Use the Eurachem equations to calculate uncertainty of calibration.
Estimate bias by analyzing some standards (with a very strong preference for ones that you did not use to generate the calibration) and comparing your result with the known CO2 content. Correct your results from the calibration by adding or subtracting the bias as appropriate.
Peter
For my high range co2 curve blanks are outside the range, so if I compile scans at 500, 800, 1000, and 1500 ppm, I should be able to plot them and extrapolate between them, correct?
And I have 13 of these curves, many of which were made with some blank scans (like the low range co2) so for these I should just be able to add in some blank scans?
If you have a calibration that is linear then you need to fit the best straight line (that is probably what your software is doing anyway) rather than interpolating between the points. If the calibration is so non-linear, and cannot be fitted by a straight line or a curve that makes analytical sense, so that you have to interpolate, you need more than four points. I would prefer to see at least one more point anyway - check the official guidance that applies to medical gas analysis.
Why are you so keen to include blanks ? The problem with blanks is that you do not know their CO2 content - you assume it is zero but if water can get in then CO2 can as well - and all signal vs response curves are non-linear at their lower ends - there is always a non-zero quantity of analyte that gives a zero (+/- noise) signal. If you want to use the data you have already then just delete the points for the blanks. Make sure you document why for when the regulators come around.
Peter
Peter, that last response I was talking about the bias, I did not actually specify in the post. I can have a reportable bias at 500 ppm and at 800 ppm, but what if my concentration ends up being 650 ppm? That is what I am talking about when I mention interpolating between points and trying to tie bias to concentration. I'm in the process of plotting bias' at different concentrations to see if they are actually linear and if I can get some sort of usable relationship.
As far as blanks, I have 13 different calibration curves for different components ranging from CO2 to CO to ethylene to freon. Some of these curves use regions that are virtually free of interference. For instance, my trichloroethylene curve has a blank average of -0.004 ppm with a standard deviation of 0.003 ppm. Those seem like totally reasonable values to include in a curve, it is just more points to use. Some of these components should not be in normal air and I am monitoring them at levels around 0.5 ppm so blank scans actually do represent values close to what I am looking at.
The CO2 example was a bad one to use for discussing blanks for the reasons you mentioned and I will not be including the blanks in those determinations.
My thoughts are for the curves that only cover a narrow range, I can probably average a few bias' to get a single usable bias value that can be applied to the entire curve. This assumes the bias values are similar across the entire curve of course. For the curves that have a wider calibrated region I am going to also plot bias' across several concentrations and hope that it gives a linear response, or a fairly constant value or % that can be applied.
I will be using the Eurochem uncertainty guide Page 19 under Bias Study.