I'm wondering if I can call the following procedure “cross validation”.
I extracted k independent sets of data with comparable size from the same population. One of them was used for model development / training and all other sets were used to validate this particular model. So there is no “rotation” where each of the k data sets take turns to act as the training set. So is this a “cross-validation”?
Just be clear on what you're doing, it's better than giving names.
I would say your case is plausible cross validation---you have a training data set and multiple validations. Others might say the contrary, doesn't matter really.
You can also wait for others to cross validate my answer, and by the way, there aren't clear training cycles here too.
I think calling it “hold-out validation” instead of “cross-validation” could avoid some confusion.
Incidentally, I would be interested in the thinking behind it. Why not simply assess prediction accuracy on all the holdout data instead of partitioning it in k samples. Also, if you have so much data in each sample that it is enough to reasonably train the model, why go to the trouble of taking additional samples instead of just taking two?
