If you do chromatographic analysis you usually work with methods where peaks are integrated and possibly background signal has been subtracted. This means no signal no integration and therefore a perfect point for your calibration will be (0,0)!? Do you agree?!
By DaT on Tuesday, May 27, 2003 - 06:06 am:
You can't use (0,0) as a calibration std. Check whatever books you got there, they'll say the same.
In theory when there are no background signals you go through (0,0).
In theory, when you subtract an excisting background signal you should go to (0,0). Yet it will be very hard to get your curve through (0,0) though. And is it even necesarry? not in my opinion.
So: in theory it is the perfect point. If your lineair curve (i assume you use a linear curve) gets to this equation: Y = bx + a then you can check easily whether for x = 0 (conc = 0) your signal (Y) can be forced through 0 or not.
Construct 95% erm dono the english word damn, erm, curves for 95% assurances that a measurement falls in that area (in dutch they are called 95% betrouwbaarheidscurves). If the point (0,0) falls in that 95% area, then you may force your linear kalibration curve through (0,0).
I dono for sure whether this is what you asked though.
I have all this in an excell file, which i can give you if you want.
By DaT on Tuesday, May 27, 2003 - 06:08 am:
i wanna add: i did a validation and my (0,0) was in that 95% area. Then i optimized the method and did the validation again. Once more the point (0,0) was in the 95% area. Still in neither cases i forced the curve through zero. In my opinion you have to stand by what you measred, even if statistically it says you can push it through zero.
By Homann on Wednesday, May 28, 2003 - 06:59 am:
Thank you for your fine answer. I agree with your conclusions, and also that you should stick to your data. However, since you twice managed to get the (0,0) within the 95% confidence level, I would also be tempted to say that this also supports my "hypothesis". I have not found any book dealing with the subject in detail i.e. explaning in detail the things with integration in chromatography, but perhaps you know of a good source? - does your excell file give some instructions?!
By DaT on Monday, June 2, 2003 - 02:18 am:
sry, was on a holiday.
excell doesn't give any instrctions at all to construct 95% curves.
I made them following the statistic books i was forced to buy at school so :s
Maybe if you search the web? or mail me, i can give you my excell file
By Labcat on Monday, June 2, 2003 - 08:23 am:
Excel doesn't construct 95% confidence limits curves around the regression line, but it can give these limits for slope and intercept, which is what you need, after all. You can obtain them with the Data Analysis ToolPak, Regression analysis.
If you want the plot of the entire curve you can use some statistical package; I know there is some shareware in the Internet, though I don't have the address at hand.
As far as I know it is never considered a good practice to insert the blank sample results in the calibration curves (Check also: Miller & Miller, Statistics for analytical chemistry).