A simpler way to think about it might be "How would I assign a p-value to the personality type?" Once you have that answer, all your "over-representation" stat

A simpler way to think about it might be "How would I assign a p-value to the personality type?" Once you have that answer, all your "over-representation" stats should, naturally, trim out (or otherwise down-weight) the poor p-value personality assignments (any such over-representation stats that you have heard very likely do no such filtering).