6:29 a.m.
Hi all,
I need an efficient means to simulate a lot of autoregressive panel data, but have not
been able to do it without a lot of trouble.
Thanks in advance,
Fred
6:56 a.m.
On Wed, 21 Oct 2020, Fred Engst wrote:
Hi all,
I need an efficient means to simulate a lot of autoregressive panel data, but have not
been able to do it without a lot of trouble.
Does this help?
<hansl>
nulldata 360
setobs 12 1:1 --stacked-time-series
series x = normal()
series y = 10
y = 1 + 0.9 * y(-1) + x - 0.9*x(-1) + normal()
dpanel 1 ; y const x x(-1) --system
</hansl>
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
9:24 a.m.
Hi all,
After I send the email out, I’m figured out one approach is to first simulate a time
series, and then recast it into a panel by using the setobs command.
However, my purpose is to see the distribution of the panel adf test, but in panel adf, I
can’t retrieve the test statistics, nor the p-value of the test using $pvalue or $test.
What should I do?
Once more, thanks in advance.
Fred
On Oct 21, 2020, at 2:29 PM, Fred Engst <engst.uibe(a)gmail.com>
wrote:
Hi all,
I need an efficient means to simulate a lot of autoregressive panel data, but have not
been able to do it without a lot of trouble.
Thanks in advance,
Fred
9:32 a.m.
On Wed, 21 Oct 2020, Fred Engst wrote:
Hi all,
After I send the email out, I’m figured out one approach is to first
simulate a time series, and then recast it into a panel by using the
setobs command.
That would make the cross-sectional units not independent. As to whether
it consitutes a problem, it depends on the context. I'm not sure what the
requirements are for the Choi meta-tests, for example.
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
10:13 a.m.
Am 21.10.2020 um 11:24 schrieb Fred Engst:
However, my purpose is to see the distribution of the panel adf test,
but in panel adf, I can’t retrieve the test statistics, nor the p-value of the test using
$pvalue or $test.
This should probably be fixed (made available after the command).
> On Oct 21, 2020, at 2:29 PM, Fred Engst
<engst.uibe(a)gmail.com> wrote:
>
> Hi all,
> I need an efficient means to simulate a lot of autoregressive panel data, but have
not been able to do it without a lot of trouble.
Not sure what "trouble" means here. I think in a panel dataset the
autoregressive simulations work just like in a single time-series,
automatically unit-per-unit. (As in Jack's example.)
cheers
sven
12:38 p.m.
On Wed, 21 Oct 2020, Fred Engst wrote:
[M]y purpose is to see the distribution of the panel adf test, but
in panel adf, I can’t retrieve the test statistics, nor the
p-value of the test using $pvalue or $test.
What should I do?
You should be able to get $test and $pvalue using the "levinlin"
command. As for "adf" itself, in the panel case as things stand
you'd have to save results by "scraping" the output, as illustrated
below.
<hansl>
function matrix panel_adf_pv (string s)
string line
matrix pv = {}
loop while getline(s, line)
lb = strstr(line, "[")
if lb != ""
pv ~= atof(lb+1)
endif
endloop
return pv
end function
open grunfeld.gdt
string s
outfile --buffer=s
adf 2 value --quiet
end outfile
matrix pvs = panel_adf_pv(s)
print pvs
levinlin 2 value --quiet
eval $test
eval $pvalue
</hansl>
Allin
1528
days inactive
1528
days old
5 comments
4 participants
participants (4)
-
Allin Cottrell -
Fred Engst -
Riccardo (Jack) Lucchetti -
Sven Schreiber