0
$\begingroup$

I used k folds cross validation to calculate the predective error of my linear mixed-effects model. I computed it manually because the caret package is not available from lmer model of lme4 package. My question is: how can summarize k folds cross validation results using someindex and some plots to understand if my model make good prediction? I calculated MSE for each folds and the following index:

enter image description here

Thanks so much

$\endgroup$
1
$\begingroup$

You can just average over the results of the k-folds. Do note that k-fold cross validation is used to see which model works best, not to judge the expected error of the final model on new data. To do the final testing, use data that you (hopefully) kept separate during the entire modelling phase.

$\endgroup$
  • $\begingroup$ Thanks @dimpol for the response. Precisely I segmented my data into 10 folds. I used iteratively the train data (all dataset without a k fold) to fit the model, and the test data (the dataset only in the k fold) to calculate prediction. Finally I calculate MSE for each iteration( k iteration). is it incorrect? $\endgroup$ – Angela Andreella Oct 20 '16 at 10:46

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.