"Although it is nice to have an unbiased estimator of the variance, we do not really need it to understand the relation between our independent variable and our dependent variable. Why?"
I think I kind of get it, but he wants a specific answer, and he picks on me...