7:53 p.m.
Dear Gretl Community,
I would like to report two "issues" about VAR estimation.
# Part 1: "--out-of-sample" problem
I'm trying to perform an out of sample forecast of a VAR model but Gretl is
telling me that I have an error with my data. Please take a look at my
script:
<hansl>
open australia.gdt
smpl ; 1990:1
"VAR" <- var 4 lpau lpus le --seasonals --quiet
dataset addobs 4
fcast 1990:2 1992:1 # "Works ok!"
fcast 1990:2 1992:1 --dynamic # "Works ok!"
fcast 1990:2 1992:1 --static # "Works ok!"
fcast 1990:2 1992:1 --out-of-sample "Does not work"
</hansl>
# Part 2: Saving VAR inside a loop problem
These one looks like a small bug. When I try to save models with quoted
names Gretl doesn't save them. Please take a look at my script:
<hansl>
open australia.gdt
loop i=1..2 # "Works ok!"
smpl 1972:1 1990:$i
VAR$i <- var 4 lpau lpus le --seasonals --quiet
endloop
smpl --full
loop i=1..2 # "Does not work"
smpl 1972:1 1990:$i
"VAR$i" <- var 4 lpau lpus le --seasonals --quiet
endloop
</hansl>
--
*Henrique C. de Andrade*
Doutorando em Economia Aplicada
Universidade Federal do Rio Grande do Sul
www.ufrgs.br/ppge
3:14 p.m.
On Fri, 29 Jul 2011, Henrique Andrade wrote:
I would like to report two "issues" about VAR estimation.
# Part 1: "--out-of-sample" problem
"VAR" <- var 4 lpau lpus le --seasonals --quiet
[...]
fcast 1990:2 1992:1 --out-of-sample "Does not work"
You're not allowed to specify dates with the --out-of-sample
option; it's a shorthand form that figures the dates for you.
# Part 2: Saving VAR inside a loop problem
These one looks like a small bug. When I try to save models with quoted
names Gretl doesn't save them. [...]
Yes, that one was a bug. Now fixed in CVS.
Allin Cottrell
4:41 p.m.
Dear Allin,
At first I would like to thank you for your help! But I'm a little
bit confused with this:
Em 30 de julho de 2011 Allin escreveu:
On Fri, 29 Jul 2011, Henrique Andrade wrote:
> I would like to report two "issues" about VAR estimation.
>
> # Part 1: "--out-of-sample" problem
>
> "VAR" <- var 4 lpau lpus le --seasonals --quiet
[...]
> fcast 1990:2 1992:1 --out-of-sample "Does not work"
You're not allowed to specify dates with the --out-of-sample
option; it's a shorthand form that figures the dates for you.
In my humble opinion this is not very intuitive. As we know,
the "--out-of-sample" option gives us static forecasts inside
the sample and dynamic one outside the sample, so, (again)
in my humble opinion sometimes an user will need perform
a forecast like that:
<hansl>
*open* australia.gdt
*smpl* ; 1990:1
"VAR" <- *var* 4 lpau lpus le *--seasonals --quiet*
*dataset* addobs 4
*fcast* 1972:1 1992:1 *--out-of-sample*
</hansl>
What do you think?
Um abraço,
*Henrique C. de Andrade*
Doutorando em Economia Aplicada
Universidade Federal do Rio Grande do Sul
www.ufrgs.br/ppge
5:14 p.m.
On Sat, 30 Jul 2011, Henrique Andrade wrote:
At first I would like to thank you for your help! But I'm a
little
bit confused with this:
Em 30 de julho de 2011 Allin escreveu:
>
> You're not allowed to specify dates with the --out-of-sample
> option; it's a shorthand form that figures the dates for you.
In my humble opinion this is not very intuitive. As we know,
the "--out-of-sample" option gives us static forecasts inside
the sample and dynamic one outside the sample,
No. What you describe is the _default_ behavior of "fcast" for
a dynamic model. Here's the help text:
"Forecasts are generated for a certain range of observations:
if startobs and endobs are given, for that range (if
possible); otherwise if the --out-of-sample option is given,
for observations following the range over which the model was
estimated."
"The default is to give a static forecast for any portion of
the forecast range that lies within the sample range over
which the model was estimated, and a dynamic forecast (if
relevant) out of sample."
That is, the --out-of-sample option is equivalent to
"fcast t1 t2" where t1 = the first observation following the
estimation range and t2 = the end of the dataset. This will be
a dynamic forecast if possible.
Allin
4895
days inactive
4896
days old
3 comments
2 participants
participants (2)
-
Allin Cottrell -
Henrique Andrade