[SemiAuto]

This topic came up at work as my supervisor usually includes or forces zero in all calibration curves.  When is using 'zero' or the 'origin' appropriate in calibration curve and when is it not?  A lot of our methods either 'include' or 'force' zero I guess the thinking that if nothing is there then there will be no response.  We use GC with TCD and FID detectors mostly, and one ECD.  We do Internal Standard and External Standard methods mostly.

Recently a pHD of ours declared that especially for assay methods including zero/origin is plainly wrong by accepted best practices.  I don't see the big deal myself because the FID methods that I have looked at are in the dynamic linear range and basically would go through zero/origin anyway. 

So forum members what say you?

[JGK]

Having worked in contract research organizations (CROs) field all of my career, I have never forced a calibration through zero or used a zero value as a calibration point.

Indeed, when working to good laboratory practice (GLP), we are not permitted to report results outside of the calibration range used (which is commonly defined as the range of non-zero STDs).

I totally agree with your PhD guy.

[ARGOS++]

Dear SemiAuto;

You should never use the Origin inside any calibration if you not have a reliable measurement for it.
But you may do it if you don’t care about your results and/or your own reliability at all.
(I met enough bad examples!)

To test with the “Blank Experiment” or “Zero Experiment” your Calibration is also an important part of any task that is called a “Real Calibration” and its Validation.
That’s precisely why the deviation of the Origin is a Quality Criterion for how good, how valid your Calibration will be, especially if you’re missing a real cause for such found deviations!

Maybe you should also read:   "Why a Blank Experiment”

Good Luck!
                    ARGOS⁺⁺

[SemiAuto]

It is not that I am necessarily expected to report below the lowest actual calibration point, but that it really doesn't seem to shift my calibration curve (line) much at all whether origin/zero is included or not.  My R= 1.0000 without zero in one recent example verses an R= 0.99995 including origin... it changed the assay by only 0.13%.  I could see how the case could be made in this particular example that it doesn't make a significant difference, but I want to be doing it right and proper myself.

[ARGOS++]

Dear SemiAuto;

It sounds not bad, but R is never a measure how good you target the origin.
R is a statistical Parameter that indicates how close your measuring points lay to a straight line (best fit), and as a statistical Parameter R is quite a strongman for systematically errors compared to what your eye is able to detect.

But back to your calibration: Important (especially for weak samples even more!) at what values your straight line will have its intersections with your X-Axis and with your Y-Axis.

Your reported deviation of 0.13% is only valid for particular Sample of a particular Concentration, but does not tell how good your calibration is over the useful range.
(But if all your errors lay only in such a close range, then it will by quite good for GC.)

I hope I have been of help to you.

Good Luck!
                    ARGOS⁺⁺

[Elric]

My momma told me never to use the word never ;0)

In reality those people who deal with low-level samples, at or near the detection limt, commonly force through zero to avoid y-intercept corrections at or equal to the sample signal.  Read EPA methods here.  Using an average of response factors by definition is forcing through zero.  So you can listen to all the scholarly information here or follow a real life example, satisfying more legal cases than you ever want to be involved in.  So let everyone give their opinion, but in the end the most litigious data in the world (billions in EPA cleanup costs at low levels of contamination) is collected, reported and defended in court using calibration data forced through zero.

Let the mud slinging begin.

Elric

[ARGOS++]

Dear Elric;

Of course!, that it’s for everybody allowed to believe what ever everybody likes, for EPA too!

But it’s at least of the same true! that every systematically error is, now and for ever!, a systematically Mistake! — absolutely independent who ever is doing it!

Using a not measured, or a forced Point in any Statistics, and Calibration is Statistics, is there wile violating the tiniest Statistics and every Logic too!

Quintessence:
Everybody who is doing it anyway, is not only disqualifying the Calibration she/he/it just did!

It’s also not with, but violating the Proverb:

“Wer misst, misst Mist!, was weiter nicht schlimm ….. . ”

from the Topic:  "Fatty Acids and UV Light”
(There are also translation hints available.)


Good Luck!
                    ARGOS⁺⁺