problems with normally distributed residuals

M

*marie*

Guest
#1
Dear all,
I hope you can help me with the following problem:

I estimate a time series, using OLS.As a typical problem autocorrelation arises. I solved this problem with Newey West Standard errors. But with one of my regressions the post testing results that my residuals are not normally distributed. Usually one can solve this by the rreg or qreg regression in stata. But what about the autocorrelation?

Can anybody tell me if the rreg and qreg regression has also robust standard errors in the autocorrelation problem?

How can I regress this correctly?

Hope anyone can help me. Many thanks in advance!
 
#2
You might want to look at

---ARIMA--- for time series. Very powerful command which can take care of autocorrelation.

Best wishes
Stephen
 
M

*marie*

Guest
#3
Dear Stephen,

many thanks for your reply. To make sure the problem is not the autocorrelation itself, but the combination with the not normally distributed residuals post regression (as I described). If you write ARIMA is useful in autocorrelation problems, I do not see the answer to my residual problem...?

I used the stata command newey y x1 x2 in my regression (this produces robust standard errors for autocorrelation)
But in the post regression test my residuals came out as non-normally distributed. Usually I would use rreg y x1 x2 (or qreg) to solve this problem, but I am not sure if this regression automatically corrects my standard errors?

Many thanks.