Page 1 of 1
Auto regressive distributed lagged models.
Posted: Wed Jan 18, 2012 4:21 pm
by nattieboat
how do I estimate auto regressive distributed lagged models in eviews. I'm applying the ARDL approach to cointegration but facing some basic difficulties. Can someone please teach me how to estimate using the accurate maximum number of lags and how the intepretation would look like. I'm using eviews 5.1.
Re: Auto regressive distributed lagged models.
Posted: Mon Jan 23, 2012 2:21 pm
by nattieboat
ISN'T THERE ANYONE WHO CAN HELP ME?
Re: Auto regressive distributed lagged models.
Posted: Tue Jan 24, 2012 1:38 am
by donihue
For equation 1, it is really very easy to estimate the equations by introducing the various lagged variables. The hard part is in determining the number of lags. There is no "formulaic" answer to that. The "general to specific" approach of the LSE school suggests starting with a maximal number of lags for each variable and then testing down by eliminating insignificant lags (you can do this easily with EViews "redundant variables" test). If you have quarterly data, typically you would start with 4 lags per variable (assuming you have enough data to avoid degrees of freedom problems)
Equation 2 is then a repeat of the work for equation 1, using differenced variables and adding in your "ecm" term.
Regards
Donihue