How does the efficiency of a regression predict the power of the t-test? How do robust standard errors change the comparison between hypothesis testing using OLS and robust regression? What does dependence in the e do to the t-test? what does non-normality in the e do to the t-test?