1:49 p.m.
Is there a way to estimate the best ARIMA model within hansl? Granted
I could just generate a half dozen or so and compare selection
criteria, but is there a more elegant way? Tks
TMN
2:01 a.m.
Thanks, I see your point.
TMN
On Sun, May 18, 2014 at 9:52 PM, Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 18.05.2014 15:49, schrieb Tim Nall:
> Is there a way to estimate the best ARIMA model within hansl? Granted
> I could just generate a half dozen or so and compare selection
> criteria, but is there a more elegant way? Tks
Define "best".
-sven
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
--
Best regards,
Timothy M. Nall
Assistant Professor
National Quemoy University
Kinmen, Taiwan
3:07 a.m.
On Mon, 19 May 2014, Tim Nall wrote:
Thanks, I see your point.
TMN
On Sun, May 18, 2014 at 9:52 PM, Sven Schreiber <svetosch(a)gmx.net> wrote:
> Am 18.05.2014 15:49, schrieb Tim Nall:
>> Is there a way to estimate the best ARIMA model within hansl? Granted
>> I could just generate a half dozen or so and compare selection
>> criteria, but is there a more elegant way? Tks
>
> Define "best".
Well, if you define "best" as the model which optimizes a chosen
information criterion, the gretl function package "armax" (by
Yi-Nung Yang) does that.
/Tools/Function packages/on server
Allin Cottrell
4:05 a.m.
I wasn't sure if armax could be accessed via hansl.... Moreover,
and slightly off-point, armax might be improved by having its
function calls preceded by "catch" logic, since it crashes frequently.
I wrote something in gretl/python that duplicates the basic
functionality of armax. Mine and armax crash at the same points --
something about being "unable to compute the Hessian", according to my
very flawed memory -- but the error can be caught and the next
iteration typically works just fine. I don't know if that's meaningful
info or not. I don't know whether continuing on after a crash is a
menangful thing to do.
Anyhow, thanks.
TMN
On Mon, May 19, 2014 at 11:07 AM, Allin Cottrell <cottrell(a)wfu.edu> wrote:
On Mon, 19 May 2014, Tim Nall wrote:
> Thanks, I see your point.
> TMN
>
> On Sun, May 18, 2014 at 9:52 PM, Sven Schreiber <svetosch(a)gmx.net> wrote:
>> Am 18.05.2014 15:49, schrieb Tim Nall:
>>> Is there a way to estimate the best ARIMA model within hansl? Granted
>>> I could just generate a half dozen or so and compare selection
>>> criteria, but is there a more elegant way? Tks
>>
>> Define "best".
Well, if you define "best" as the model which optimizes a chosen
information criterion, the gretl function package "armax" (by
Yi-Nung Yang) does that.
/Tools/Function packages/on server
Allin Cottrell
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
--
Best regards,
Timothy M. Nall
Assistant Professor
National Quemoy University
Kinmen, Taiwan
4:13 a.m.
oops I meant "hansl/python," not "gretl/python". It's just a
little
hansl script, and then I physically copy/paste its output to a text
file and process its text via python. Nothing special.
On Mon, May 19, 2014 at 12:05 PM, Tim Nall <tnall.ling(a)gmail.com> wrote:
I wasn't sure if armax could be accessed via hansl....
Moreover,
and slightly off-point, armax might be improved by having its
function calls preceded by "catch" logic, since it crashes frequently.
I wrote something in gretl/python that duplicates the basic
functionality of armax. Mine and armax crash at the same points --
something about being "unable to compute the Hessian", according to my
very flawed memory -- but the error can be caught and the next
iteration typically works just fine. I don't know if that's meaningful
info or not. I don't know whether continuing on after a crash is a
menangful thing to do.
Anyhow, thanks.
TMN
On Mon, May 19, 2014 at 11:07 AM, Allin Cottrell <cottrell(a)wfu.edu> wrote:
> On Mon, 19 May 2014, Tim Nall wrote:
>
>> Thanks, I see your point.
>> TMN
>>
>> On Sun, May 18, 2014 at 9:52 PM, Sven Schreiber <svetosch(a)gmx.net> wrote:
>>> Am 18.05.2014 15:49, schrieb Tim Nall:
>>>> Is there a way to estimate the best ARIMA model within hansl? Granted
>>>> I could just generate a half dozen or so and compare selection
>>>> criteria, but is there a more elegant way? Tks
>>>
>>> Define "best".
>
> Well, if you define "best" as the model which optimizes a chosen
> information criterion, the gretl function package "armax" (by
> Yi-Nung Yang) does that.
>
> /Tools/Function packages/on server
>
> Allin Cottrell
> _______________________________________________
> Gretl-users mailing list
> Gretl-users(a)lists.wfu.edu
> http://lists.wfu.edu/mailman/listinfo/gretl-users
--
Best regards,
Timothy M. Nall
Assistant Professor
National Quemoy University
Kinmen, Taiwan
--
Best regards,
Timothy M. Nall
Assistant Professor
National Quemoy University
Kinmen, Taiwan
11:25 p.m.
On Mon, 19 May 2014, Tim Nall wrote:
I wasn't sure if armax could be accessed via hansl...
All function packages can be accessed via hansl. Just use the
"include" command to access the .gfn (package) file.
Moreover, and slightly off-point, armax might be improved by
having its function calls preceded by "catch" logic, since it
crashes frequently.
Please reserve the word "crash" to refer to a condition in which
gretl itself exits prematurely (generally due to a [memory]
segmentation fault). What you're talking about with armax we would
describe as an "error condition": gretl doesn't quit, but the
execution of the script is halted.
That said, your suggestion that "catch" should be used is a good
one.
I don't know whether continuing on after a crash is a menangful
thing to do.
After a true crash there's no option to continue: the OS causes the
program to exit. After a script error it's certainly possible to
continue, and it will be meaningful to do so unless the script in
question unconditionally assumes that no error has occurred, in
which case the errors may compound and become increasingly hard to
understand.
Allin Cottrell
3876
days inactive
3877
days old
6 comments
3 participants
participants (3)
-
Allin Cottrell -
Sven Schreiber -
Tim Nall