3:33 p.m.
Hello,
It's really lacking some tests when you estimate a panel model. Could it be
added some panel unit root tests, for example, please?
Best wishes,
Bruno
3:56 p.m.
Bruno Thiago Tomio schrieb:
Hello,
It's really lacking some tests when you estimate a panel model. Could it
be added some panel unit root tests, for example, please?
In principle you're right, but I guess it's a barrel without bottom (old
German proverb). What test in particular would you be interested in? BTW
the earlier panel unit root tests have ridiculously strong independence
assumptions IMHO. Also this may be a good case for contributed function
packages (I'm not 100% sure if it's really that convenient for panel
data, though.)
cheers,
sven
4:17 p.m.
New subject: RES: Panel models tests
Right... I'm aware of some limitations of these kind of tests. Anyway, tests
like IPS, Maddala-Wu or Levin-Lin-Chou (2002) would be welcome. I was hoping
to find a R package for these tests, but no success. Last but not least,
thank you for answering, Sven.
Sincerely,
Bruno
-----Mensagem original-----
De: gretl-users-bounces(a)lists.wfu.edu
[mailto:gretl-users-bounces@lists.wfu.edu] Em nome de Sven Schreiber
Enviada em: quarta-feira, 6 de janeiro de 2010 13:57
Para: Gretl list
Assunto: Re: [Gretl-users] Panel models tests
Bruno Thiago Tomio schrieb:
Hello,
It's really lacking some tests when you estimate a panel model. Could
it be added some panel unit root tests, for example, please?
In principle you're right, but I guess it's a barrel without bottom (old
German proverb). What test in particular would you be interested in? BTW the
earlier panel unit root tests have ridiculously strong independence
assumptions IMHO. Also this may be a good case for contributed function
packages (I'm not 100% sure if it's really that convenient for panel data,
though.)
cheers,
sven
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
11:23 p.m.
New subject: RES: Panel models tests
Well I don't have a complete solution for you, but AFAIR some of these
tests rely on average ADF test statistics. So here is an example script
showing how to calculate such averages across panel units:
open greene14_1.gdt # example of a panel dataset
# which variable?
series tested = C
lagorder = 0
scalar numofunits = max($unit)
matrix tstats = {}
loop for i=1..numofunits
smpl $unit=i --restrict
adf lagorder tested --c
matrix tstats = tstats | $test
smpl full
endloop
tbar = meanc(tstats)
print tbar
For averaging p-values instead just replace $test with $pvalue. After
some polishing (and generalization) all this stuff could be wrapped in
function packages, but currently I don't have time for that, sorry.
Anybody else?
cheers,
sven
Bruno Thiago Tomio schrieb:
Right... I'm aware of some limitations of these kind of tests.
Anyway, tests
like IPS, Maddala-Wu or Levin-Lin-Chou (2002) would be welcome. I was hoping
to find a R package for these tests, but no success. Last but not least,
thank you for answering, Sven.
Sincerely,
Bruno
-----Mensagem original-----
De: gretl-users-bounces(a)lists.wfu.edu
[mailto:gretl-users-bounces@lists.wfu.edu] Em nome de Sven Schreiber
Enviada em: quarta-feira, 6 de janeiro de 2010 13:57
Para: Gretl list
Assunto: Re: [Gretl-users] Panel models tests
Bruno Thiago Tomio schrieb:
> Hello,
>
> It's really lacking some tests when you estimate a panel model. Could
> it be added some panel unit root tests, for example, please?
>
In principle you're right, but I guess it's a barrel without bottom (old
German proverb). What test in particular would you be interested in? BTW the
earlier panel unit root tests have ridiculously strong independence
assumptions IMHO. Also this may be a good case for contributed function
packages (I'm not 100% sure if it's really that convenient for panel data,
though.)
cheers,
sven
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
2:06 p.m.
New subject: RES: RES: Panel models tests
Thank you very much, Sven. I get "Data error" when trying to run tbar =
meanc(tstats), following your instructions.
Cheers,
Bruno
-----Mensagem original-----
De: gretl-users-bounces(a)lists.wfu.edu
[mailto:gretl-users-bounces@lists.wfu.edu] Em nome de Sven Schreiber
Enviada em: sexta-feira, 8 de janeiro de 2010 21:24
Para: Gretl list
Assunto: Re: [Gretl-users] RES: Panel models tests
Well I don't have a complete solution for you, but AFAIR some of these tests
rely on average ADF test statistics. So here is an example script showing
how to calculate such averages across panel units:
open greene14_1.gdt # example of a panel dataset
# which variable?
series tested = C
lagorder = 0
scalar numofunits = max($unit)
matrix tstats = {}
loop for i=1..numofunits
smpl $unit=i --restrict
adf lagorder tested --c
matrix tstats = tstats | $test
smpl full
endloop
tbar = meanc(tstats)
print tbar
For averaging p-values instead just replace $test with $pvalue. After some
polishing (and generalization) all this stuff could be wrapped in function
packages, but currently I don't have time for that, sorry.
Anybody else?
cheers,
sven
Bruno Thiago Tomio schrieb:
Right... I'm aware of some limitations of these kind of tests.
Anyway,
tests like IPS, Maddala-Wu or Levin-Lin-Chou (2002) would be welcome.
I was hoping to find a R package for these tests, but no success. Last
but not least, thank you for answering, Sven.
Sincerely,
Bruno
-----Mensagem original-----
De: gretl-users-bounces(a)lists.wfu.edu
[mailto:gretl-users-bounces@lists.wfu.edu] Em nome de Sven Schreiber
Enviada em: quarta-feira, 6 de janeiro de 2010 13:57
Para: Gretl list
Assunto: Re: [Gretl-users] Panel models tests
Bruno Thiago Tomio schrieb:
> Hello,
>
> It's really lacking some tests when you estimate a panel model. Could
> it be added some panel unit root tests, for example, please?
>
In principle you're right, but I guess it's a barrel without bottom
(old German proverb). What test in particular would you be interested
in? BTW the earlier panel unit root tests have ridiculously strong
independence assumptions IMHO. Also this may be a good case for
contributed function packages (I'm not 100% sure if it's really that
convenient for panel data,
though.)
cheers,
sven
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
2:28 p.m.
New subject: RES: RES: Panel models tests
Bruno Thiago Tomio schrieb:
Thank you very much, Sven. I get "Data error" when trying
to run tbar =
meanc(tstats), following your instructions.
Hm, I'm guessing maybe some missings in your dataset?
Does the script run ok for you with the example dataset (shipped with
gretl)? If not, what gretl version are you using?
-sven
4:19 p.m.
New subject: RES: RES: RES: Panel models tests
Using greene14_1.gdt, the script output is:
gretl version 1.8.6
Current session: 2010-01-11 14:17
? scalar numofunits = max($unit)
Generated scalar numofunits = 6
? matrix tstats = {}
Generated matrix tstats
? loop for i=1..numofunits
smpl $unit=i --restrict
adf lagorder tested --c
matrix tstats = tstats | $test
smpl full
endloop
loop: i = 1
? smpl $unit=i --restrict
Full data set: 90 observations
Current sample: 15 observations
adf: expected an integer order
> adf lagorder tested --c
-----Mensagem original-----
De: gretl-users-bounces(a)lists.wfu.edu
[mailto:gretl-users-bounces@lists.wfu.edu] Em nome de Sven Schreiber
Enviada em: segunda-feira, 11 de janeiro de 2010 12:29
Para: Gretl list
Assunto: Re: [Gretl-users] RES: RES: Panel models tests
Bruno Thiago Tomio schrieb:
Thank you very much, Sven. I get "Data error" when trying
to run tbar
= meanc(tstats), following your instructions.
Hm, I'm guessing maybe some missings in your dataset?
Does the script run ok for you with the example dataset (shipped with
gretl)? If not, what gretl version are you using?
-sven
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
9:30 p.m.
New subject: RES: RES: RES: Panel models tests
Bruno Thiago Tomio schrieb:
Using greene14_1.gdt, the script output is:
gretl version 1.8.6
Current session: 2010-01-11 14:17
? scalar numofunits = max($unit)
Generated scalar numofunits = 6
? matrix tstats = {}
Generated matrix tstats
? loop for i=1..numofunits
> smpl $unit=i --restrict
> adf lagorder tested --c
> matrix tstats = tstats | $test
> smpl full
> endloop
loop: i = 1
? smpl $unit=i --restrict
Full data set: 90 observations
Current sample: 15 observations
adf: expected an integer order
>> adf lagorder tested --c
You need to run the entire script, I'm repeating it here, this time in
tags (which I usually add, but this time I left them out because I was
starting to think they look awkward, but maybe not...):
<script>
open greene14_1.gdt # example of a panel dataset
# which variable?
series tested = C
lagorder = 0
scalar numofunits = max($unit)
matrix tstats = {}
loop for i=1..numofunits
smpl $unit=i --restrict
adf lagorder tested --c
matrix tstats = tstats | $test
smpl full
endloop
tbar = meanc(tstats)
print tbar
</script>
-sven
9:47 p.m.
New subject: RES: RES: RES: RES: Panel models tests
It was my bad, Sven. Sorry and thanks for your help.
Bruno
-----Mensagem original-----
De: gretl-users-bounces(a)lists.wfu.edu
[mailto:gretl-users-bounces@lists.wfu.edu] Em nome de Sven Schreiber
Enviada em: segunda-feira, 11 de janeiro de 2010 19:30
Para: Gretl list
Assunto: Re: [Gretl-users] RES: RES: RES: Panel models tests
Bruno Thiago Tomio schrieb:
Using greene14_1.gdt, the script output is:
gretl version 1.8.6
Current session: 2010-01-11 14:17
? scalar numofunits = max($unit)
Generated scalar numofunits = 6
? matrix tstats = {}
Generated matrix tstats
? loop for i=1..numofunits
> smpl $unit=i --restrict
> adf lagorder tested --c
> matrix tstats = tstats | $test
> smpl full
> endloop
loop: i = 1
? smpl $unit=i --restrict
Full data set: 90 observations
Current sample: 15 observations
adf: expected an integer order
>> adf lagorder tested --c
You need to run the entire script, I'm repeating it here, this time in tags
(which I usually add, but this time I left them out because I was starting
to think they look awkward, but maybe not...):
<script>
open greene14_1.gdt # example of a panel dataset
# which variable?
series tested = C
lagorder = 0
scalar numofunits = max($unit)
matrix tstats = {}
loop for i=1..numofunits
smpl $unit=i --restrict
adf lagorder tested --c
matrix tstats = tstats | $test
smpl full
endloop
tbar = meanc(tstats)
print tbar
</script>
-sven
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
4:40 p.m.
New subject: RES: RES: Panel models tests
sven
I run your script on gretl 1.8.6 (under winXP) and it works great.
Yi-Nung Yang
2010/1/11 Sven Schreiber <svetosch(a)gmx.net>:
Bruno Thiago Tomio schrieb:
> Thank you very much, Sven. I get "Data error" when trying to run tbar =
> meanc(tstats), following your instructions.
>
Hm, I'm guessing maybe some missings in your dataset?
Does the script run ok for you with the example dataset (shipped with
gretl)? If not, what gretl version are you using?
-sven
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
5459
days inactive
5464
days old
9 comments
3 participants
participants (3)
-
Bruno Thiago Tomio -
Sven Schreiber -
yinung＠Gmail