Page 2 of 2
Re: multicollinearity concerns panel data across section or
Posted: Thu Nov 13, 2014 9:43 am
by startz
I misunderstood. Using a lagged dependent variable may be okay, but how do you know that it eliminates serial correlation?
Re: multicollinearity concerns panel data across section or
Posted: Thu Nov 13, 2014 9:46 am
by Wouter van der Stee
Durbin Watson, or is that not a valid test if a lagged independent variable in included on teh right hand side? If so, is Durbin Watson a valid test for ar(1) included?
Re: multicollinearity concerns panel data across section or
Posted: Thu Nov 13, 2014 9:51 am
by startz
Durbin-Watson is not a valid test in this circumstance. Use the Breusch-Godfrey test under View/Residual Diagnostics/Serial Correlation and set to 1 lag
Re: multicollinearity concerns panel data across section or
Posted: Thu Nov 13, 2014 9:59 am
by Wouter van der Stee
That's not possible when running a GARCH(1,1) model. I use Eviews 7. The only test available is the ARCH LM test under this window...
Re: multicollinearity concerns panel data across section or
Posted: Thu Nov 13, 2014 11:10 am
by startz
That's not possible when running a GARCH(1,1) model. I use Eviews 7. The only test available is the ARCH LM test under this window...
I'm not sure what to suggest as a work around, as the DW can be wildly off in this circumstance. My guess is that you should stick with using AR(1).
Re: multicollinearity concerns panel data across section or
Posted: Fri Nov 14, 2014 1:27 am
by Wouter van der Stee
Startz,
Thank you very much for your time! Helped me a lot.
Wouter