Hi! I want to use an Augmented Engle-Granger cointegratoin approach to ARDL, where the VEC contain first differences form from the ARDL, since I'll loose valuable simultaneous year correlation with ordinary VEC. Now, how many lags in the Augmented Engle-Granger cointegratoin should I use? Also, can I just include dummy varibales in the original regression, from which I build the Augmented Engle-Granger cointegratoin, in order to account for structural breaks?
This is a paper on Augmented Engle-Granger cointegratoin approach to ARDL: https://www.google.se/search?q=LAG+LENGTH+SPECIFICATION+IN+ENGLE-GRANGER+COINTEGRATION+TEST:+A+MODIFIED+KOYCK+MEAN+LAG+APPROACH+BASED+ON+PARTIAL+CORRELATION&ie=utf-8&oe=utf-8&gws_rd=cr&ei=-qkmVt-PAYKfsAHM-L-4Cw
Hi, this may help: How to perform ARDL by using microfit,
http://phd-inceif.blogspot.my/search/label/ARDL
You should add lags until you remove the possible autocorrelation. If I had, say, 50 annual data, I would include up to 5 lags and then, I would remove them until the last one were significantly different from 0 (t-rat>1.65). You can also use some information criterion. Use the paper of Perron and Qu (2007) in this regard.
Hi, you can conduct the ARDL cointegration approach by eviews:
http://davegiles.blogspot.com/2015/01/ardl-modelling-in-eviews-9.html
The Engle-Granger cointegration approach is a two-step procedure that involves estimating an unrestricted autoregressive model for each of the variables in the system and then testing for cointegration using the residuals from these regressions. The approach was proposed by Robert Engle and Clive Granger in 1987.
The ARDL model is a generalization of the Engle-Granger approach to cointegration. It allows for the inclusion of both stationary and non-stationary variables in the same regression equation
- Badges
- Science topic