Chromatography Forum

Dear All

I've come across this equation (in a PhD thesis from 1991) for calculating an LOD from a calibration plot. Despite a thorough search, I have not been unable to find any published articles which have used this, so no idea where it came from. Can anybody help me?

Sy/x = ((Sum of (yi'-yi)2)/n-2)raised to the power 1/2

where yi = observed peak height (or analyte/internal std ratio)
and yi' = predicted peak height (or ratio) and n= number of points in calibration curve

Many thanks in advance

Ann

Sy.x normally refers to the standard deviation of the individual deviations of measured values, in Y-units. You can find these formulas in a booklet 'Use of statistics to develop and evaluate analytical methods', Grant T Wernimont, AOAC 1985, ISBN 0-935584-31-5

regards Bert

Ann,
I'm sure I have seen the equation:
LOD = intercept + (3 x blank standard deviation)
This is a good book:
Ref:
http://www.amazon.co.uk/exec/obidos/ASI ... 17-8051853

Also:
Ref:
http://www.chem.utoronto.ca/coursenotes ... Regr2a.pdf

On page 7.

Hope this helps
WK

Dear Bert and WK

Many thanks for your replies, much appreciated. I have seen the equation using blank samples but I really need to reference the equation in my first post as this was used for the determination of LOD in some work I did a few years ago (not the way I would do things now but we live and learn ).

Bert, is the equation I used actually in the book you recommended (I'll certainly try to get hold of that book but as it's out of print it may take me a while)?

Once again, thank you for taking the time to respond.

All best wishes, Ann

Dear Ann,

what I wanted to say is that I think it is a strange interpretation of the term Syx. I never have seen that kind of calculation for the LOD and to be honoust, I think it is not correct. Sorry

Good luck. I would be very interested to hear if you find any reference.

regards Bert

Thanks for your help Bert, I'll certainly post again if I find the source!

Very best regards, Ann

Is the issue that you have to calculate LOD using existing standard curve data, and don't have access to sufficient blanks to determine noise? If so, there is a formula in ISO 11843-2 to calculate minumum detectable value x[sub]d using linear standards. It's on p 5 of ISO 11843-2, or p 317 of "Statistical Methods for Quality Control, Vol 2," ISO 2000. I'm not going to try to reproduce the formulas here, but they do resemble the formula that started this discussion.

Thank you for your reply Mary, I shall try to get hold of those references.

All best wishes, Ann

I assume that LOD is referring to Limit of Detection. In this case, I have always used the ICH guidelines of LOD being when the analyte peak yields 3 times baseline noise, and LOQ is when the analyte peak is 10 times baseline noise.

Is this approach too simplistic?