[evanb]

or

3) there's a misunderstanding about ordinary least-squares.