4:03 p.m.
We're getting ready to unveil a new built-in estimator, namely
random effects probit (you use "probit" command with the
--random-effects option). A question arises: where should this
go under the /Model menu in the main gretl window?
It could go under Panel, or Limited dependent variable (or
possibly both, though that may not be a very happy precedent).
I guess my own preference would be to make random effects an
option in the model specification dialog that appears if you
select
/Model/Limited dependent variable/Probit/Binary
(with panel data only, of course). But we thought we'd ask if
anyone has strong feelings about this.
(My feelings are not very strong, but what I suggest above has
the attraction of mirroring the situation on the command
line.)
Allin
8:03 p.m.
Am 03.06.2013 18:03, schrieb Allin Cottrell:
We're getting ready to unveil a new built-in estimator, namely
random effects probit (you use "probit" command with the
--random-effects option). A question arises: where should this
go under the /Model menu in the main gretl window?
It could go under Panel, or Limited dependent variable (or
possibly both, though that may not be a very happy precedent).
I think this option could be worth exploring... Of course we do not want
to clutter the menus too much, but OTOH, if it logically fits under
both, then any single place will involve arbitrariness. So what are the
downsides in the sense of: would that precedent mean that suddenly a
bunch of other menu entries would logically have to be multiplied? Or is
it really an isolated case (so far)?
cheers,
sven
12:15 a.m.
On Mon, 3 Jun 2013, Sven Schreiber wrote:
Am 03.06.2013 18:03, schrieb Allin Cottrell:
> We're getting ready to unveil a new built-in estimator, namely
> random effects probit (you use "probit" command with the
> --random-effects option). A question arises: where should this
> go under the /Model menu in the main gretl window?
>
> It could go under Panel, or Limited dependent variable (or
> possibly both, though that may not be a very happy precedent).
I think this option could be worth exploring... Of course we do not want
to clutter the menus too much, but OTOH, if it logically fits under
both, then any single place will involve arbitrariness. So what are the
downsides in the sense of: would that precedent mean that suddenly a
bunch of other menu entries would logically have to be multiplied? Or is
it really an isolated case (so far)?
I believe it's an isolated case so far. The possible downside I see
(leaving aside the question of precedence) is that users might
wonder if the two menu entries did the same thing, given that the
default presumption is (I would suppose) one item on a given menu
per command-line possibility.
Allin
12:46 a.m.
If you were to pick one, I think it should be under panel estimators since
otherwise it will (or should) be greyed out if you do not have a panel data
set declared.
My choice would be to see it both places in the GUI since its purpose is to
make the program easier to use. Some will look for it in binary others in
panel. Where it lies within the file structure of the underlying program
should not drive where it shows up in the GUI which is meant for a
different type of user.
Sadly, many will use anything for which there is a button. There is bound
to be misuse of anything statistical as all of us can attest.
Cheers,
Lee
On Mon, Jun 10, 2013 at 7:15 PM, Allin Cottrell <cottrell(a)wfu.edu> wrote:
On Mon, 3 Jun 2013, Sven Schreiber wrote:
> Am 03.06.2013 18:03, schrieb Allin Cottrell:
>> We're getting ready to unveil a new built-in estimator, namely
>> random effects probit (you use "probit" command with the
>> --random-effects option). A question arises: where should this
>> go under the /Model menu in the main gretl window?
>>
>> It could go under Panel, or Limited dependent variable (or
>> possibly both, though that may not be a very happy precedent).
>
> I think this option could be worth exploring... Of course we do not want
> to clutter the menus too much, but OTOH, if it logically fits under
> both, then any single place will involve arbitrariness. So what are the
> downsides in the sense of: would that precedent mean that suddenly a
> bunch of other menu entries would logically have to be multiplied? Or is
> it really an isolated case (so far)?
I believe it's an isolated case so far. The possible downside I see
(leaving aside the question of precedence) is that users might
wonder if the two menu entries did the same thing, given that the
default presumption is (I would suppose) one item on a given menu
per command-line possibility.
Allin
_______________________________________________
Gretl-devel mailing list
Gretl-devel(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-devel
--
Lee Adkins
Professor of Economics
lee.adkins(a)okstate.edu
learneconometrics.com
1:27 a.m.
On Mon, 10 Jun 2013, Lee Adkins wrote:
(re. placement of a GUI menu item for random effects probit, under
Limited dependent variable models or under Panel models, or both)
If you were to pick one, I think it should be under panel estimators
since
otherwise it will (or should) be greyed out if you do not have a panel data
set declared.
I take your point. Per my original suggestion, however, it would be
an option that appears in the dialog box when binary probit is
selected -- and that option would appear only if the dataset were
in panel form.
My choice would be to see it both places in the GUI since its purpose
is to
make the program easier to use. Some will look for it in binary others in
panel. Where it lies within the file structure of the underlying program
should not drive where it shows up in the GUI which is meant for a
different type of user.
OK, that looks like two votes for the double placement of the random
effects probit item (Sven and Lee) with no votes for other options.
Any other thoughts before we proceed?
Allin
4220
days inactive
4228
days old
4 comments
3 participants
participants (3)
-
Allin Cottrell -
Lee Adkins -
Sven Schreiber