ECON321 Lecture Notes - Lecture 11: Homoscedasticity, Bias Of An Estimator, Econometrics
Document Summary
We found that the variance of our ols estimators is: Xij xj 2 1 rj because we do not know the ui, we cannot calculate. In the case of k+1 parameters, df=n-k-1: if the errors are homoskedastic, then an estimate of the var( ) using 2 is unbiased. Introduction to econometrics, notes 11 p. 2: gauss markov theorem the ols estimator is the best linear unbiased estimator (blue) 1 is the partial effect of x1 on y. So then 1 is just our estimate of the partial effect of x1 on y. But this interpretation of , and of , is not accurate in all cases. Suppose we estimate the following regression model: ln wi = 0 1educi 2experi 3experi. 1 is the partial effect of education on ln(w), if we increase education by 1 year, we believe ln(w) should increase by 1 units. It is not the partial effect of work experience on log wages!