Each row of the table lists the values of the predictors for a terminal
node. When the analysis uses cross-validation or no validation, the fit for the
row is for the tree fit with the entire dataset. If you use a test data set,
the fit is from the cases in the training data.