Page 1 of 1
Eliminate Autocorrelation with a Method other than Ar(1)
Posted: Sat Sep 01, 2012 10:52 am
by yoyoo
Hi,
I have used ar(1) to address the issue of autocorrelation. The Durbin-Watson value reached 1.85 which is excellent. However, none of the independent variables become significant. Is there other than ar(1) to eliminate autocorrelation? Please see attached.
Thank you in advance.
Re: Eliminate Autocorrelation with a Method other than Ar(1)
Posted: Mon Sep 17, 2012 11:39 am
by yoyoo
Dear All,
Can anyone help me on this? If not, could you please refer me to where I can find a solution.
Thank you in advance.
Re: Eliminate Autocorrelation with a Method other than Ar(1)
Posted: Mon Sep 17, 2012 12:31 pm
by EViews Glenn
You can add lagged endogenous to the specification to whiten the residuals.
Re: Eliminate Autocorrelation with a Method other than Ar(1)
Posted: Mon Sep 17, 2012 4:09 pm
by startz
You can add lagged endogenous to the specification to whiten the residuals.
But don't do it unless you have a very specific reason to. AR(1) and a lagged endogenous variable are not the same specification and lead to very different interpretations of the dynamic response of he system to the exogenous variables.
Re: Eliminate Autocorrelation with a Method other than Ar(1)
Posted: Tue Sep 18, 2012 1:32 pm
by EViews Glenn
True, unless there are no other regressors. It didn't seem to me that the original AR specification was chosen with a particular dynamic structure in mind. I was going to add originally, but didn't, that if the significance of results changed a lot, I would worry...