N5

Consider the regression model y, = Bo + B, > + Bixa + u. Let p, denote the OLS swimator of By and let B. = > wy. where the was may be functions of the Wins and Xos. Even if the Ait has, and ys are obtained by random sampling there is no perfect collinearity, E(u, (x x, ) = ( and varu >, x,) = 0'. we can find Wa such that F(B ) - B, and var(B.)