Limit of detection and error in analysis

[DocHollywood]

Hey Guys,
I've searched the site to get an idea of how to correctly determine the detection limits and introduce error bars on my data. Unfortunately i've encountered a few problems.
For example
This guys states that:

"from the linearity curve find out residual standard deviation and slope.
Now one can derive

LOD = 3.3*residual stdandard deviation / slope.

LOQ = 10*residual stdandard deviation / slope.

FOR finding out resdiual standard deviation one can use MS-EXCEL application. for the selected two sets of variables follow =styex(Y1:Yn, X1:Xn)."

To calibrate the GC i've run two standards. 1) at 1% methane/ He balance and a 15 ppm methane/ He balance standard. Since i plot the concentration vs peak area, my slope is around 0.0003x. I ofcourse cannot get the residual standard deviation as well since i have only two calibration standards. Does anyone have a good technique to figure out error bars in this type of situation. I think for LOD i will take the 3 times the noise level and for LOQ i will take 10 times the noise level.
Thanks

[Peter Apps]

Run 5 replicates at each level, then work out the relative standard deviation at each level. If the rsd is less than 0.33 and 0.1 you can claim that the LOD and LOQ respectively are below that level. Further than that you cannot rigourously go - there is no way to do statisitics on a two-point "calibration".

Peter

[carlos.teixeira]

Dear DocHollywood,
I am curious, did you solve your challenge?

Have you nice elutions!!

Carlos Teixeira

[larkl]

Ok, I do the LOD completely differently. This was shown to me years ago, when I was a co-op student.

I print out a peak with a known concentration. I determine the peak height manually using a ruler and the y axis scale.

Then I print out the same part of the chromatogram with no peak. Then I determine the noise in this area (again with a ruler and the Y axis scale).

Then LOD = 3 X Noise height X known peak conc / known peak height

[DocHollywood]

Thanks for all the responses.
I've decided to purchase some more low ppm methane standards. The problem I have is that when i use mass flow controllers as i get to higher dilutions, my error starts going way up. So currently i have to standards, one fairly high concentration and one low. I'd like to get error bars on my data but i still haven't figured out a good way to do that.
Thanks again

[Peter Apps]

Error bars show the standard deviation (usually) - to determine that you need to run replicates at each level.

Peter

[DocHollywood]

@ Peter,
I can get the standard deviation on my calibration points. Thats not the problem. The problem is how do I relate that to error bars on my data. So for example if i have 3 calibration point 15,100,1000 ppm methane (taken a number of times to lower the STD for each point) and I have four data points 5, 50, 100 and 150 ppm. I have yet to find a good method to display the error in the data points.
Thanks for the help
AD