• You may wonder when you should use robust standard errors and when you should
use ordinary OLS standard errors.
          • The advantage of OLS standard errors is that t-tests using them are exact tests, as long
as the assumptions of the classical linear regression model are satisfied.
          • So OLS standard errors are better if you believe these assumptions are reasonable.
          • Robust standard errors are good if you have a large data set and you do not want to
rely on assumptions regarding the form of heteroskedasticity present.
          • On the other hand, if you have a small data set and you know heteroskedasticity is a
problem, your best option is to use WLS and transform the model to one where the
CLM assumptions hold.
           In a word,small data set ordinary OLS,big dada set Robust.