In data models, what is the reason for child datasets? Would it not be easier to just create a root dataset with no children?