9:42 p.m.
Dear Allin,
I'd found some strings not marked for translation (Main menu -> Help ->
Check for updates):
"You are running gretl 1.8.6.
This is the current version."
Best,
Henrique C. de Andrade
Doutorando em Economia Aplicada
Universidade Federal do Rio Grande do Sul
www.ufrgs.br/ppge
9:56 p.m.
On Fri, 22 Jan 2010, Henrique wrote:
I'd found some strings not marked for translation (Main menu
-> Help ->
Check for updates):
"You are running gretl 1.8.6.
This is the current version."
The thing is that it would require some redesign to make that
translatable: the string is issued by the server, and is only
printed by gretl.
Allin.
10:20 p.m.
Em 22 de janeiro de 2010 Allin escreveu:
On Fri, 22 Jan 2010, Henrique wrote:
> I'd found some strings not marked for translation (Main menu -> Help ->
> Check for updates):
>
> "You are running gretl 1.8.6.
> This is the current version."
The thing is that it would require some redesign to make that
translatable: the string is issued by the server, and is only
printed by gretl.
Ok! It's not a big deal ;)
Best,
Henrique
4:21 a.m.
New subject: exponential MA smoothing filter
I think there is some bug in the exponential MA filter
of gretl (Variable/filter/exponential MA).
I tried it with the Ct data in data3-6.gdt. If I ask for the filter with
the default options, including "weight on current observation = 0.100"
the graph produced is very strange (a parabola) and if I ask to save the
series, in the description of the saved series appears the text
"Exponential moving average of Ct (current weight 0.9)".
I tried writing 0.9 in the "weight on current observation" and the
resulting graph is rather more logic, but the series seems to be placed
one period ahead.
--
Ignacio Diaz-Emparanza
DEPARTAMENTO DE ECONOMÍA APLICADA III (ECONOMETRÍA Y ESTADÍSTICA)
UPV/EHU
Avda. Lehendakari Aguirre, 83 | 48015 BILBAO
T.: +34 946013732 | F.: +34 946013754
www.ea3.ehu.es
4:42 a.m.
New subject: exponential MA smoothing filter
On Tue, 26 Jan 2010, Ignacio Diaz-Emparanza wrote:
I think there is some bug in the exponential MA filter
of gretl (Variable/filter/exponential MA).
I tried it with the Ct data in data3-6.gdt. If I ask for the filter with
the default options, including "weight on current observation = 0.100"
the graph produced is very strange (a parabola) and if I ask to save the
series, in the description of the saved series appears the text
"Exponential moving average of Ct (current weight 0.9)".
I tried writing 0.9 in the "weight on current observation" and the
resulting graph is rather more logic, but the series seems to be placed
one period ahead.
If you try the following
<script>
open data3-6.gdt
series foo = mean(Ct)
series foo = 0.9*foo(-1) + 0.1*Ct
</script>
and compare foo with the automatically-generated series, you'll see two
things: one, you're right about the time displacement. Two, the parabola
effect is a consequence of initialising the EWMA with the sample mean,
which of course produces strange results with a trending series such as Ct
in your example. Do you have a better idea?
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti
5:29 a.m.
New subject: exponential MA smoothing filter
El mar, 26-01-2010 a las 10:42 +0100, Riccardo (Jack) Lucchetti
escribió:
If you try the following
<script>
open data3-6.gdt
series foo = mean(Ct)
series foo = 0.9*foo(-1) + 0.1*Ct
</script>
and compare foo with the automatically-generated series, you'll see two
things: one, you're right about the time displacement. Two, the parabola
effect is a consequence of initialising the EWMA with the sample mean,
which of course produces strange results with a trending series such as Ct
in your example. Do you have a better idea?
Not really. We have to put something in the default and I suppose the
mean of the whole series is a correct posibility for stationary data.
I was trying to replicate the smoothed series in Granger's (1989) book
"Forecasting in Business and Economics" (table 2.2 in page 25). In page
27 he proposed the filter (x_t is the original series, y_t is the
filtered series)
y_1=x_1, y_t=k y_t-1 + (1-k)x_t, t=2,...n.
"The larger the value of k, the greater the smoothing achieved. Typical
appropiate values for k would be 0.7 for monthly data, 0.5 for quarterly
data, and 0.3 for a series recorded annually. The smoothed consumption
series, with k=0.3 has been shown in table 2.2".
I can see that the "Weight on current observation" in the gretl dialog
box corresponds to (1-k), and setting this to 0.7 and the tickmark on
"the first EMA value is" pointing to "the mean of the first n
observations" (with n=1), I obtain exactly the same values in the book
but with a displacement one period ahead.
--
Ignacio Diaz-Emparanza
DEPARTAMENTO DE ECONOMÍA APLICADA III (ECONOMETRÍA Y ESTADÍSTICA)
UPV/EHU
Avda. Lehendakari Aguirre, 83 | 48015 BILBAO
T.: +34 946013732 | F.: +34 946013754
www.ea3.ehu.es
7:16 a.m.
New subject: exponential MA smoothing filter
On Tue, 26 Jan 2010, Ignacio Diaz-Emparanza wrote:
Not really. We have to put something in the default and I suppose
the
mean of the whole series is a correct posibility for stationary data.
I just did a CVS commit that ought to fix the time displacement bug.
Allin, could you have a look when you have time?
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti
9:28 a.m.
New subject: exponential MA smoothing filter
On Tue, 26 Jan 2010, Riccardo (Jack) Lucchetti wrote:
On Tue, 26 Jan 2010, Ignacio Diaz-Emparanza wrote:
> Not really. We have to put something in the default and I suppose the
> mean of the whole series is a correct posibility for stationary data.
I just did a CVS commit that ought to fix the time displacement bug.
Allin, could you have a look when you have time?
The original coding of EMA was confusing, sorry. I've tried to
clarify it in CVS; please check that I haven't screwed it up.
Allin.
10:58 a.m.
New subject: exponential MA smoothing filter
El mar, 26-01-2010 a las 09:28 -0500, Allin Cottrell escribió:
On Tue, 26 Jan 2010, Riccardo (Jack) Lucchetti wrote:
> On Tue, 26 Jan 2010, Ignacio Diaz-Emparanza wrote:
>
> > Not really. We have to put something in the default and I suppose the
> > mean of the whole series is a correct posibility for stationary data.
>
> I just did a CVS commit that ought to fix the time displacement bug.
> Allin, could you have a look when you have time?
The original coding of EMA was confusing, sorry. I've tried to
clarify it in CVS; please check that I haven't screwed it up.
I think it is OK now. But please could you correct the descriptive
labels of the ema_ and emc_ resulting series? The value of "current
weight" reported is one minus the value you write in the dialog box, and
I think this should be the same value.
--
Ignacio Diaz-Emparanza
DEPARTAMENTO DE ECONOMÍA APLICADA III (ECONOMETRÍA Y ESTADÍSTICA)
UPV/EHU
Avda. Lehendakari Aguirre, 83 | 48015 BILBAO
T.: +34 946013732 | F.: +34 946013754
www.ea3.ehu.es
11:38 a.m.
New subject: exponential MA smoothing filter
On Tue, 26 Jan 2010, Ignacio Diaz-Emparanza wrote:
El mar, 26-01-2010 a las 09:28 -0500, Allin Cottrell escribió:
> On Tue, 26 Jan 2010, Riccardo (Jack) Lucchetti wrote:
>
> > On Tue, 26 Jan 2010, Ignacio Diaz-Emparanza wrote:
> >
> > > Not really. We have to put something in the default and I suppose the
> > > mean of the whole series is a correct posibility for stationary data.
> >
> > I just did a CVS commit that ought to fix the time displacement bug.
> > Allin, could you have a look when you have time?
>
> The original coding of EMA was confusing, sorry. I've tried to
> clarify it in CVS; please check that I haven't screwed it up.
I think it is OK now. But please could you correct the descriptive
labels of the ema_ and emc_ resulting series? The value of "current
weight" reported is one minus the value you write in the dialog box, and
I think this should be the same value.
OK, I see what you mean. That's now fixed in CVS.
Allin.
5409
days inactive
5413
days old
10 comments
5 participants
participants (5)
-
Allin Cottrell -
Henrique -
Ignacio Diaz-Emparanza -
Riccardo (Jack) Lucchetti -
Sven Schreiber