Whit least squares calibration, data variability is usually accessed in order to decide for OLS (unweighed) or WLS (weighted) least squares approaches.
Parameter and concentration estimation is clear in several published works.
But, what about other validation parameters such as detection and quantification limits?
Dear Jorge,
Check this paper out and also the reference. The authors are here on RG:
Article Weighted Least-Squares Approach To Calculating Limits of Det...
Detection is well done graphically in a graphical residual analysis. This can be carried forward for complete treatment of weighted least squares (WLS) regression. Please see https://www.researchgate.net/project/OLS-Regression-Should-Not-Be-a-Default-for-WLS-Regression
The following part of that project may be of particular interest:
https://www.researchgate.net/publication/333642828_Estimating_the_Coefficient_of_Heteroscedasticity
Please note that OLS regression is just a special case of WLS regression. It is generally easier to implement (possible exception: the classical ratio estimator), but often OLS is not a good choice.
- Badges
- Science topic
- Similar topics
- Psychology