Re: [Gretl-devel] smpl <condition> --restrict --permanent
On Thu, 2 Jul 2015, Sven Schreiber wrote:
Am 02.07.2015 um 01:46 schrieb Allin Cottrell:
>
> Any comments/suggestions?
Previously I forgot to add one remark: What about the treatment of pre-sample
values with lagged terms? It seems that models can still lose their
underlying data if the sample restriction is treated literally. This can be
tricky.
Actually, this is automatically taken care of by my proposal. The
"subsample" I'm talking about, in relation to saved models, is the set
of observations from the full dataset that is currently "available" or
"visible". The observations used in estimation (including any
pre-sample lags) must always be a subset (proper or not) of this set.
While gretl will sometimes access values prior to the lower limit of
the sample "range", $t1 (if a model includes lags and $t1 > 1), it
will never access values outside of what I'm calling the "subsample",
as set by a boolean restriction.
Allin