Chromatography Forum

I have come across some articles saying that both " adding point 0,0 " and "force to origin" are bad practices in constructing a calibration curve (y=mx+c).

I also think that in spectrophotometric methods such as UV, Fluorescene and IR , matrix effect or instrumental effect will always give signals (+ive or -ive) when no analyte is present.

Adding a non-experimental point 0,0 or "force to origin"
alters the real result from the experiment and make the cal. curve biased from the real experimental condition.

However, in some situation like chromatographic method , e.g. GC-FID

My thought is that :

two instrumental situation should be corrected before conducting GC-FID experiment:

1, no analyte but signal (or Peak) present at the retention time of the analyte
2, negative signal occur at the retention time of the analyte when no analyte presents

that means GC-FID system should be error (or abnormal situation) free (at least the errors of above) before doing experiment.

Then we use the above error-free system to calibrate curve using y=mx+c method.

However, if (1) the calibration curve generates x-value > zero (e.g. amount) when y-value (e.g. peak area) equals zero,
(2) the calibration curve generates negative x-value

Is it contradicting with the instrumental correction above ?

Adding point (0,0) to the curve will not solve the problem of -ive x-value , so only "force to origin" solves both -ive x-value problem and "zero analyte giving zero calculated amount" problem.

Does the "Force to origin " treatment make the cal. curve more closer to the real experimental condition ?

Thanks all for the reply !

Does the "Force to origin " treatment make the cal. curve more closer to the real experimental condition ?

I think it doesn't.

When one observes Y values different than 0, when X=0, there has to be same systematic error that produces y-intercept in the calibration diagram.
Using the least square method to construct the "real" calibration curve gives the possibility to find the best slope for the curve that fits the experimental results.

So if your least square cal. curve doesn't pass through the center (0,0) that means that you are probably dealing with systematic error, which can arise from usage of bad solvents, analyst mistakes during sample preparation, instrumental errors (bad signal from the detector due to hardware or software problem) e.t.c.
Finding the source of that systematic error is, I think, probably the best approach to minimize or eliminate the factors that produce errors (y-intercept on the calibration diagram).

Forcing something "always to pass through some point" e.g. (0,0) biases the experimental results and doesn't represent the actual situation/or the actual relationship between the independent variable X and the dependent variable Y.

Best regards

The answer depends a bit on how large your non-zero intercept is.

Most statistics packages (and even Microsoft Excel) will calculate the standard error of the y-intercept. If that interval (y intercept +/- the standard error) includes zero, then it is statistically indistinguishable from zero. If the intercept is large enough that your interval does not include zero, then as zokitano said, "forcing" zero is, in general, a bad idea.

In my opinion one should not do any chromatography without frequent checks for carryover. If one does this one has plenty of data showing where the zero injection point lies. No need to discuss "forcing through zero".

In my opinion one should not do any chromatography without frequent checks for carryover. If one does this one has plenty of data showing where the zero injection point lies. No need to discuss "forcing through zero".

Exactly!
When there are always present residual analytes with every further injection, or carryover, one should not expect (if other preparation/instrument parameters are good) Y-intercept to be 0. This (carryover) is also one of the potential sources of systematic error.

Including for e.g. needle wash function minimizes the carryover, thus chromatographic information gained are more accurate represents of the relationship between the mass (concentration) of the analyte and the signal measured.