Consider the multiple regression model = + , where () = 0, but () = ^2, where V is a known n*n positive definite symmetric matrix. Note this means the errors are not homoscedastic (in the classic regression we assume that V = I, the identity matrix.) Let _GLS denote the generalized least-squares estimator of , defined as _GLS= argmin ( )' ^(-1) ( ) It can be shown that _GLS = ('^(-1))^(-1)'^(-1), you can use this result without proof. a) Prove that _GLS is unbiased for and (_GLS) = ^2 ('^(-1))^(-1)