[Gretl-users] Re: Troubles with residuals recovery in a VAR
Am 13.10.2020 um 16:38 schrieb Allin Cottrell:
On Tue, 13 Oct 2020, Daniel Ventosa-Santaulària wrote:
> Hello everyone,
> I am having trouble with the recovery of the residuals in a VAR model. If
> if do it though the GUI everything is ok, but if I try to do it through a
> code, then I can only recover the residuals (as a time series) of the
> first equation, but for the remaining equations, I can only get them as a
> matrix. Here is this simple example code:
>
> open np.gdt
> var 2 rgnp emply interest --robust
> series Resid1=$uhat[,1] # This one works perfectly
> series Resid2=$uhat[,2] # This one does not; mismatch between series and
> matrix
>
> this is probably a rookie's mistake, but I cannot see where the
> problem is.
It seems like this ought to work, I need to take a closer look.
Actually, as per the user guide section 17.10 this is not officially
supported. (Since the $uhat matrix/vector is shorter than the current
sample.) I guess you (Allin) are referring to some internal storage of
t1 and t2 as matrix properties which would align it to the relevant
subsample range. My guess is that this somehow gets lost after the 1st
column. If this is supposed to be official (and then fixed), I guess it
should be documented somehow.
thanks
sven