[Gretl-devel] Re: Simulating a simultaneous system
Am 07.10.2020 um 13:45 schrieb Allin Cottrell:
On Wed, 7 Oct 2020, Sven Schreiber wrote:
> Am 06.10.2020 um 21:53 schrieb Allin Cottrell:
>> The first use of fcast does not exactly reproduce Sim2; that's because
>> the two leading rows are static forecasts rather than actual observed
>> values.
>
> Not sure I understand what you mean here?
In the Sim2 produced by varsimul the first two rows are just original
data, the simulation as such begins on row 3. But in fcast with
starting date $nobs - 1 all rows of the output are "predictions":
plain fitted values for the first 2 then dynamic forecasts for the
remainder. The respective dynamic forecasts will therefore not be the
same.
Ah OK, right. Of course it's not obvious why you would want to start
forecasting already in $nobs-1 (BTW, I really prefer to use $t2 because
$nobs depends on the start of the sample as well, but of course it's
equivalent here...) and then have such a mish-mash of static and dynamic
forecasts. I guess you wanted to match the number of rows from varsimul
which includes the last observed as starting values. Let me just note
that while the default behavior of fcast may not be the most natural
here, all this and also the varsimul behavior is properly documented and
the user has full control already, so all is good.
thanks
sven