confidence interval in calibration

[lanospam]

Dear All,

I perform a spectrometric (infrared) calibration to predict protein content
in food. Let's say that my calibration error (mean square error between ir
predicted values and reference data) is 0.2. I also know that my reference
method for protein dosage has alsoo its own error, lets say 0.05.
If I have to compute the total error of my calibration method in order to
build a confidence interval for true protein content, how can I mix these
two errors. I guess it's not just adding the variance, is it ?

Thank you for your help.

Fred

[rindtr]

Post by lanospam
Dear All,
I perform a spectrometric (infrared) calibration to predict protein content  
in food. Let's say that my calibration error (mean square error between ir
predicted values and reference data) is 0.2. I also know that my reference
method for protein dosage has alsoo its own error, lets say 0.05.
If I have to compute the total error of my calibration method in order to
build a confidence interval for true protein content, how can I mix these
two errors. I guess it's not just adding the variance, is it ?
Thank you for your help.
Fred  

hi there,
I would like to ask whether your error stationary or not it means ,is
that 0,2 error same for all concentration.if it is so you should add
it to error which comes from your method. or it is better to calculate
uncertainty of measurement by using recoveries for at least 5
concentration and 3 parallel measurement then at the end all your
errors would be in this uncertainty of measurement range so you can
ignore all errors by giving result within this range. I hope it will
help
best wishes
Rind

[Marvin]

Post by lanospam
Dear All,
I perform a spectrometric (infrared) calibration to predict protein content
in food. Let's say that my calibration error (mean square error between ir
predicted values and reference data) is 0.2. I also know that my reference
method for protein dosage has alsoo its own error, lets say 0.05.
If I have to compute the total error of my calibration method in order to
build a confidence interval for true protein content, how can I mix these
two errors. I guess it's not just adding the variance, is it ?
Thank you for your help.
Fred

I discussed this topic in the following publication:
Fitting of Analytical Functions with Digital Computers in
Spectrochemical Analysis.
M. Margoshes and S. D. Rasberry
Anal. Chem. 41, 1163-72 (1969)

One can assume that the calibration standards are like the
samples. (If they aren't, the validity of an analysis based
on the standards is open to question.) The scatter of the
measurements around the analytical curve can be sued as an
estimate of the uncertainty that takes into account the
uncertainty related to the calibration - a factor that is
usually not taken into account.

If the reference method was used to establish the standards,
then the errors due to that method are automatically
included in the uncertainty as given by my method.

Marvin Margoshes