I have a data set that is contaminated with outliers, so to overcome the problem I applied robust regression using MM estimation, but after fitting the model using the robust regression I checked for heteroscedasticity and autocorrelation using White-test and Durbin Watson, the tests indicate the presence of heteroscedasticity and autocorrelation? should I be concerned?
I know that the robust regression methods are robust against the deviation from the normality assumption, but what about autocorrelation and heteroscedasticity?
Hi,
Usually, the heteroscedasticity is caused by the autocorrelation of error terms. So in my point of view, you need to fix the autocorrelation problem at first. There are some remedial approaches for autocorrelated errors. See for example, John Neter et. al. book, "Applied linear regression models", section 13.4.
Regards.
I agree with Dr Al Zaim. I attached a paper with an example in the last section. Best wishes David Booth
Robust regression methods are designed to be less sensitive to violations of OLS assumptions. Therefore, the assumptions of homoscedasticity and auto-correlation are not required to be checked.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis