Dear Researchers,
Anyone can indicate me how to calculate root mean squared error of cross-validation (leave-one-out) for interpolated data (RMSECVi), root mean squared error of cross-validation (leave-one-out) (RMSECV) and root mean squared error of calibration (RMSEC), using Microsoft office Excel, even if there are some formulas but still need some clarification and interpretation of them.
I am using HPLC retention time for prediction of toxicity LD50 by quantitative retention activity relationship method , From a qualitative point of view, it is said large differences between RMSEC, RMSECV and RMSECVi indicate a lack of robustness of the QRAR models obtained and the need for greater cautions in future predictions.
thanks for your help,
Serge
Dear Serge
Some information you can find in this paper: https://pdfs.semanticscholar.org/4eb2/5ff5a87f2fd6789c5b9954eddddfd1c59dab.pdf
and guidance on page 60 (using equation about total sum of squares )
http://www.oecd.org/officialdocuments/publicdisplaydocumentpdf/?doclanguage=en&cote=env/jm/mono(2007)2
or from paper by Paola Gramatica
Article On the Development and Validation of QSAR Models
Hope it helps
Best regards
Tomasz
Hey, Mr. Serge:
It may depend on the software you are using.
Herein is the method by Matlab
RMSEcv=sqrt(sum((Xi-Yi).^2)/n);
Xi: Exp.Conc
Yi: Cal.Conc
Cal values according to cross validation
n : training set number
Xi and Yi values in columns.
Best regards
- Badges
- Science method