training- validation- test of a model

[progster]

Dear all, I am using decision trees in spss, the possibilities for validation are crossvalidation or split-sample validation.

So this is for the part of training and validation datasets. for instance 70% vs 30% of cases in a dataset.

I'm a little bit confused about the theorical recomendation of splitting the data into 3 dataset, training- validation-test.

how SPSS is supposed to deal with a test dataset?