[mercury]

What is the explanation behind sensitivity of the intrument and the results we obtain from it?If the intruments calibrates on high sensitivity, are we expecting high or low bias results?What is the relationship between sensitivity and the result of samples obtained?

[Mr Peanut]

Sensitivity refers to the lowest concentration at which the technique is an effective measurement method. It is the concept behind the parameter of minimum detection limit (mdl).

Measurements made at or near the mdl are inherently imprecise. In fact, one common method for calculating the mdl is based on precision measurements at low levels (see 40 CFR 136 appendices).

Bias refers to a method's (or a method's implementation's) tendency to shift results above or below "actual" or "true" value. The concept is not necessarily related to sensitivity.


Sensitivity and "result" have an ambiguous relationship but the relationship between sensitivity and the result's precision and/or accuracy can be addressed, but only when the specifics of the analysis are presented to the addressor.

[mercury]

In that case for an example,if the calibration for my GC gives LOW detection results for the calibration points(25ppm-500ppm)(in this case would be "peak area")..how will this affect my results of samples or Quality Control result?I need explanation on what's the significance in obtaining Low or High calibration sensitivity results.

[Mr Peanut]

mercury,

What do you mean by "LOW detection results"? If you mean the areas obtained for the lower calibration points were lower than they should have been then your calibration will be in error.

If your calibration function is a line then the line will be in error. If it is a curve then the lower end will be a curve.

You see, there are too many ifs here.

[mercury]

"low detection area" meaning in each calibration point gives a low area value (this shows the GC is working on low sensitivity right?) As for the calibration line, I still get a straight line graph because all the points are in the Low bias area values. It will not produce any error. For an example:

25ppm=300 (area value)
50ppm=600
150ppm=1800
200ppm=2400

but, the week after that, after re-calibration,my results are
25ppm=500 (area value)
50ppm=1000
150ppm=3000
200ppm=4000

As you can see, both calibrations with give a nice straight line... so how are this two calibrations affecting my results precision and quality control or spike sample results? You know any websites that explain this?

[Mr Peanut]

OK, I understand now, your overall response has deteriorated. The question is why. If there is a leak then precision and accuracy could be a problem. But you say you get a straight line for the calibration indicating, perhaps, a dirty detector or a change in detector settings which has reduced sensitivity. These changes shouldn’t adversely affect accuracy or precision provided all of your readings are in the linear range defined by the calibration. (Always beware of readings near the lower end of the line because they may have a low signal to noise ratio.) To be sure you can always perform a precision and accuracy study.

I know of no website that goes into calibration troubleshooting effectively.

[Mr Peanut]

Sorry, I had to run out. But I'm back now.

You see, if your loss of sensitivity is consistent it will be the same for all standards, samples, blanks, and QC samples. So, again, the QC sample results will not be biased. Something is going on though and you should look into it.

If your applications are environmental in nature there are some guidelines you can use in qualifying data. see:
http://yosemite.epa.gov/r10/oea.nsf/Quality+Assurance/Main

.