The RATS Software Forum

Hi Tom, I am trying to implement conditional forecasting using a VAR model ("varmodel") with exogenous regressors. I augmented "varmodel" with a separate equation for the exogenous regressor. I was able to do the unconditional or the basic forecasting using the forecast command. But I can't run the conditional forecast routines. Any hint or help will be greatly appreciated.

There's nothing about the @condition procedure that requires that the model be a symmetrical VAR. I'm not sure how you intend to handle the "exogenous" variable. What @condition does assume is that shocks to all equations are available for making the restrictions hold, that is, a GDP constraint might best be explained with a shock to your exogenous variable. If that's what you intend, then you need to make sure you have a combined model with a combined (N+1) x (N+1) covariance matrix.

Right now the condition is imposed as follows:
@condition(model=varmodel+oileq,from=endate+1,steps=nstep,results=cforecasts) 1
# y(2) endate+1 .2

The exogenous regressor is oil which is specified as "oileq". Could you help me specify the variance-covariance matrix?

I assume that you'd want to create a joint covariance matrix from the residuals from all the equations (VCV instruction).

Thanks Tom. This seems to work.