For the measurement error model in (8-14) and (8-15b), prove that when only x is measured with error, the squared correlation between y and x is less than that between y* and x*. (Note the assumption that y* = y.) Does the same hold true if y* is also measured with error?