I am comparing forecast from DSGE and DSGE-VAR. For DSGE-VAR I am computing this manually.
My data has a non-zero average, so for DSGE I used option prefilter = 1
, and also for my DSGE-VAR.
I noted that when I don´t set prefilter, the estimation of DSGE-VAR is slower that when I set prefilter = 1
.
I view in Del Negro and Schorfheide (2004) that there are \Phi_{0} as the intercept term of VAR.
I am getting PHI_tilde
with size rho \times n, where n is the number of observables and rho is the number of lags. I get this size either when I set prefilter = 1
or when I do not. That is, I suspect dynare asumes there is not intercept.
So my question are:
- ¿ Is it recommended demean data series before estimation DGSE-VAR in dynare?
- ¿ Does DSGE-VAR support the prefilter-option?