I don't know what you mean by 'home factor' but my idea would be to take children from similar backgrounds who are homeschooled or not and see if that variable correlates with some measure of performance. It's not perfect of course because homeschooling parents are self-selecting, but a fully controlled experiment is not going to be feasible; however, this does not make the hypothesis any less likely, just less testable.