In the data I want to perform VAR analysis, the appropriate delay length appears to be zero. I cannot select the appropriate delay length from zero in the Eviews program. What should I do ? Can you help with this ?
Dear Hatice,
What is the data frequency in your research? If the data are annual than the zero-lag length could be appropriate. But, the following is important: the information criterions you are used are primarily of informative nature - you should not follow them too strictly. Also, do the causality test. I would recommend you to go to the end with your model and then check the residuals distribution properties. Then try with one lag and compare the suitability of the models. Of course, in your paper, you should explain the reasons for choosing some particular time lag.
I hope that it will be helpful.
Good luck with your research.
Vladimir
Hi! I would like to add just a little comment to Vladimir's answer, which is a very complete list of good ideas. Hatice's questions imply some kind of misunderstanding about what a VAR model is. Hatice's Eviews program MUST begin with delay length (lag) of 1, then 2, 3, until 12 (say) with monthly data. BUT it cannot begin with a zero lag because the implied VAR model collapses to a model that has no dynamics and supposedly no explanatory variables. Such a 'model' will 'explain' your P endogenous variables only with P stochastic disturbances. This is why Eviews is not allowing you to set a zero lag!
- Badges
- Science topic
- Similar topics
- Counseling Psychology
- Confusion