if the residuals from a fitted regression violate the assumption of homoscedasticity, we know that a. they are normally distributed. b. they are independent of one another. c. there are extreme outliers. d. their variance is not constant.