VAR(1) model - Residuals

[saraphdnz]

Hello there,

I need some guidance regarding the VAR(1) model. I have six dependent variables and I need to use VAR(1) process to obtain the residuals for each of them and then use GARCH to forecast the conditional variance based on these residuals.
My question is that what difference it makes if I select all six variables together in VAR to make the residuals 'vs' select one at a time to obtain the VAR residuals?

Regards,
Sara

[EViews Gareth]

Not sure what you mean by this. Could you explain further?

[startz]

A VAR is just a least squares regression. So as long as you use the same specification it doesn't matter if the equations are estimated together or separately.

[saraphdnz]

Thanks for the reply.

I need the residuals for the eight dependent variables (stationary as well) via VAR and I can obtain them via Eviews selecting all together as dependent variables but the residuals are showing auto-correlation. although I selected the lag length as per SIC (via lag structure).
what should I need to do here? estimate them together or one by one? is the sequence of dependent variable also important in VAR?

Please advise.

Regards,
Sara

[startz]

Thanks for the reply.

I need the residuals for the eight dependent variables (stationary as well) via VAR and I can obtain them via Eviews selecting all together as dependent variables but the residuals are showing auto-correlation. although I selected the lag length as per SIC (via lag structure).
what should I need to do here? estimate them together or one by one? is the sequence of dependent variable also important in VAR?

Please advise.

Regards,
Sara

If you are doing a standard VAR then the order of the dependent variables does not matter and whether you do them individually or together doesn't matter. If you are still seeing autocorrelation you might add another lag.

[saraphdnz]

Thanks, heaps for the explanation it is a big help.


Regards, Sara