[Gretl-devel] smpl <condition> --restrict --permanent

Allin Cottrell cottrell at wfu.edu
Thu Jul 2 10:02:03 EDT 2015


On Thu, 2 Jul 2015, Sven Schreiber wrote:

> Am 02.07.2015 um 01:46 schrieb Allin Cottrell:
>
>> So here's my proposal: in the circumstances described above, if
>> you issue a command to shrink the dataset permanently, gretl checks to
>> see whether all saved models fall under the "easy case". If so, fine. If
>> not, you get a warning that not all saved models can be preserved, with
>> a "Go ahead? Yes/No" dialog. If you say "Yes", we
>> destroy all the models that would turn into zombies.
>>
>> Any comments/suggestions?
>
> I don't remember the discussion leading to "--permanent" (hopefully it
> wasn't my request...),

As I recall, the point was: If I want to shift (fully) to a restricted 
version of my dataset, why do I have to apply the restriction, save 
the restricted dataset, then reopen the restricted version (hence 
losing my existing "session")? Why can't I just apply a permanent 
restriction on the fly?

> but anyway. In principle the idea sounds good. A couple of 
> questions:
>
> First, what happens if you do --permanent via script? Do you still 
> get a dialog?

Yes, you should (not now, but per the proposal). In the same way that 
you get a dialog if you run a script that opens a new dataset from 
file, when there's a dataset already in place.

> I guess not, but what happens instead? I guess there would be a
> case to halt with an error. However, this would also suggest the
> possibility to save the current session (to .gretl) via script commands,
> something which currently isn't possible AFAIK.

That might be worth adding anyway.

> Secondly, the current status quo of the session right before --permanent
> could be saved "automatically" in a kind of snapshot .gretl session
> file. In essence, this would be a step towards a kind of autosave
> feature which may be nice in any case.

I took Lee's point (at the gretl conference) about autosave for the 
gretl script editor. I'm not so sure about autosave for gretl 
sessions.

The proposed dialog, instead of offering just Yes/No for going
ahead with a model-destroying sample restriction, could offer a 
three-way choice:

* Back off, don't do the restriction
* Go ahead, never mind the models
* Save the current session then go ahead

> Third, why actually is --permanent not available for regular sample
> selection (without --restrict etc.)? It says so in the docs, and in the
> GUI only the corresponding dialog(s?) have that option, but I don't see
> the reason right now.

OK, that's debatable, but it would add extra complications. A defence 
of the existing situation is that it's easier to see a use-case for
wanting a "subsampling" restriction to be made permanent on the fly 
(as opposed to a t1, t2 setting) since in the former case (only) the 
dataset is basically duplicated in memory, which can become a problem 
with big data.

> Finally, what does for example Eviews do? They also have
> equation/system/model objects. Perhaps they even disallow to throw away
> the data then. I could check if that were interesting.

Yes, that might be interesting.

Allin


More information about the Gretl-devel mailing list