VAR estimation with parameter constraints

[Caleb]

Hallo altogether!

I estimated a VAR model with six variables and three time lags. The t-statistics are indicating that several parameters can't be distighished differnt from zero. To get a narrower confidence intervall for the impulse response analysis and forecasting, I'd like to restrict several parameters to zero.
I can do this with "Proc -> Make Sytem". I delete some coefficients and then estimate the model. From now on the results are presented quite different compared to the usual VAR-presentation. There are e.g. no figures for the whole system like: Determinant resid covariance (dof adj.), Determinant resid covariance, Log likelihood, Akaike information criterion and Schwarz criterion.
Further on there are no such important VAR-tools like LM-Test, AR Roots, Granger Causality, Impulse-Response, Variance Decomposition and several others.

The question is now, if I need to chose an other way for coefficient constraints or if EViews isn't able to provide these VAR-Tools with restrictions. I would be very thankful for your advice!

Best regards!
Caleb

[QSnakecharmer]

Another possible solution to estimate a VAR with parameter constraints in Eviews is to estimate the original (unrestricted) VAR model and then use the “make system” option in the “procs” option of the VAR model object. This creates a “System” object, where you could erase the parameters (variables) you don’t want to use in the estimation. It’s easy, but you have to keep in mind that because this “erased-in-some sections” model would NOT be a VAR anymore, you have to use SUR to estimate the model (also available in the System object :D ).

[abhinabaroy09]

Apart from the drop down menu. Can anyone send me the program for choosing a suitable VAR model.
Its urgent.
please mail the program @ abhinabaroy09@gmail.com.
Thank you.