Page 1 of 1

DW Statistic

Posted: Tue Feb 26, 2013 5:13 am
by TDeval
Hi there,

As previously mentioned I am conducting long horizon regressions to test the predictability of stock markets, however I am a total EViews Novice.
I have now run my regressions but was wondering if you could quickly clear up something for me.

How can I tell whether there is positive or negative autocorrelation?? I know the DW stat informs me if there is serial correlation, but how can I determine if it's positive or negative!?

Thanks you very much for your help and sorry for my stupidity.
Tom

Re: DW Statistic

Posted: Tue Feb 26, 2013 7:44 am
by startz
The Durbin-Watson is less than 2 for positive serial correlation and greater than 2 for negative serial correlation. Remember that the DW is not valid if there is a lagged dependent variable.