In the setting of $k$-fold cross validation, my data is split into $k$ disjoint sub-samples. Each fold contains one of these sub-samples as the test set $T$, and the rest of the data not in this sub-sample is assigned to the training. The model is fit on the training set, and used to predict whatever value of interest on the test set. Let's call these set of predictions $(\hat{y_i})_{i\in T}$.
I have heard of people refer to the cross validated error, that is $(\hat{y_i} - y_i)_{i\in T}$. However, can you refer to the predictions $(\hat{y_i})_{i\in T}$ as cross validated predictions? If not, what would be an appropriate way of referring to these?
