EViews.com

Dear all

I am trying to run a GARCH model with more than one explanatory variables but I am facing a lot of problems and doubts.

First, i would like to know how can we properly decide the ARCH and GARCH values to use in the model specification. I have run a GARCH(1,1) and looked to the Variance Equation. Just by looking at that I opt to use a GARCH(2,1) model, but I have no idea if this is the right thing to do.

Secondly, when performing the residuals tests I face a serious problem with auto-correlation. The results of all tests can be seen below:

• -Correlogram Q-Statistics:

• -Correlogram Squared Residuals:

• Histogram Normality Test:

• ARCH LM Test - ARCH

How can I solve this problem? I have been told (in a two minute conversation) that I should run the GARCH(q,p) model with the right value of 'q' and 'p', then do something that would generate some garch values, and then run a VAR modelusing the generated GARCH values. However, looking on the internet for this, I can't find anything that may resemble this.

Can anyone help me?

I appreciate all the help in advance.

Yours sincerely,
Afonso Rodrigues