I have been trying to calculate the root mean squares error of calibration (RMSEC) and the root mean squares error of cross validation (RMSECV) for a PCA model made in R using the mdatools package.

My dataset is too complex to upload, however I followed the same method as shown in the link below:


Am I right in saying that both RMSEC and RMSECV are calculated as the square root of the residuals^2 divided by the number of samples? From the cross validated model, I am given values for Q distance. Are these the data that I use as the residuals when calculating the RMSEC and RMSECV? If so, I would have Q residuals for the calibration model, and Q residuals for the cross validated model.

Thanks in advance


I'm trying to figure this out too, and it seems that RMSECV is related to PRESS as $$ RMSECV = \sqrt\frac{PRESS}{n*p} $$ Where PRESS is the sum of Q-residuals of each sample against a cross validated model, n is your number of samples, and p, the number variables in the original dataset

Similarly, RESS is the sum of the Q-Residuals for each sample of the model against itself, so I imagine RMSEC is similarly defined using the model Q-residuals.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.