Consider the following population model Vi = Pixi + Eiwhere & has mean 0 and variance . Note that we do not have the intercept here (it is not necessary and adding it in does not change anything). Define b, to be the least squares estimator of B1. Let b, be some other (hypothetical) estimator of B, that may not be the least squares estimator. We want to show that n n I (1) : - ( vi - b 1 x ; ) 2