This function permits the calculation of an error model from (i) a training set,
and (ii) a caret model trained on this set
to predict the response variable of interest using cross-validation (point prediction model).
The cross-validation predictions are extracted from the point prediction model.
The errors in prediction for the cross-validation predictions are then calculated.
These errors which serve as the response variable
for the error model (i.e. the error model predicts errors in prediction).
The error model uses as descriptors the same descriptors used to train the point prediction model.
These descriptors are input to the function ErrorModel through the argument "x.train".