"Statistical inference The regression model Estimating the model Assumptions on the independence of and X Hypothesis testing Assumptions on the distribution of When the assumptions break down Background reading Assumptions on the distribution of The OLS model assumes that the i are i.i.d., so E () = σ2 In×n (24) The OLS estimator is unbiased even if this "sphericality" assumption is violated, e.g., the errors are not independent or σ2 σ2 for some i, j. i j However, the OLS estimator may not have the lowest variance, among linear unbiased estimators; in the lingo, it is not the best linear unbiased estimator (BLUE)."