11:31 a.m.
Hello everybody,
I tried to build my first gretl function with no much sucess. The exercise is
very simple: write a function to obtain the theoretical simple correlations
for an AR(2) process, depending only on two parameters phi1 and phi2.
The following script works ok:
---------------
nulldata 20
series y = 1.0
scalar phi1 = 0.4
scalar phi2 = 0.5
scalar rho1 = phi1 / (1-phi2)
genr y[2] = rho1
genr y = phi1*y(-1)+phi2*y(-2)
genr y1 = y(1)
------------
So the series y1 contains the theoretical correlations.
I wrote the following script, based on the information of the manual, but the
function does not work:
--------
nulldata 20
# function definition
function corrAR2 (scalar phi1, scalar phi2)
series y = 1.0
scalar rho1 = phi1 / (1-phi2)
genr y[2] = rho1
genr y = phi1*y(-1)+phi2*y(-2)
genr y2 = y(1)
return series y2
end function
# function call
(y1) = corrAR2 0.4 0.5
--------
Which is the problem?
I tried it with official gretl-1.5.0 in mandrivalinux, and also with the
current Windows snapshot. There were no errors or warnings and I obtained
only a series with zeroes.
--
Ignacio Díaz-Emparanza
Dpto. de Economía Aplicada III (Econometría y Estadística)
UPV-EHU
12:06 p.m.
On Wed, February 8, 2006 12:31, Ignacio Díaz-Emparanza wrote:
--------
nulldata 20
# function definition
function corrAR2 (scalar phi1, scalar phi2)
series y = 1.0
scalar rho1 = phi1 / (1-phi2)
genr y[2] = rho1
genr y = phi1*y(-1)+phi2*y(-2)
genr y2 = y(1)
return series y2
end function
# function call
(y1) = corrAR2 0.4 0.5
--------
Which is the problem?
I see no big problem. Your script runs unmodified on my box. Here's the output:
jack@ec-4:~/Prove$ gretlcli Ignacio.inp
gretl version 1.5.0
Copyright Ramu Ramanathan and Allin Cottrell
This is free software with ABSOLUTELY NO WARRANTY
Current session: 2006/02/08 12:56
"help" gives a list of commands
run Ignacio.inp
/home/jack/Prove/Ignacio.inp opened OK
? nulldata 20
periodicity: 1, maxobs: 20,
observations range: 1-20
? function corrAR2 (scalar phi1, scalar phi2)
? (y1) = corrAR2 0.4 0.5
? series y = 1.0
Generated vector y (ID 4)
? scalar rho1 = phi1 / (1-phi2)
Generated scalar rho1 (ID 5) = 0.8
? genr y[2] = rho1
Modified vector y (ID 4)
? genr y = phi1*y(-1)+phi2*y(-2)
Replaced vector y (ID 4)
? genr y2 = y(1)
Generated vector y2 (ID 6)
?
and, to check that everything's ok, if I issue manually "print y1", I get
? print y1
print y1
Varname: y1
Full data range: 1 - 20 (n = 20)
Current sample: 1 - 20
0.800000 0.820000 0.728000 0.701200 0.644480 0.608392 0.565597
0.530435 0.494972 0.463206 0.432769 0.404711 0.378269 0.353663
0.330599 0.309071 0.288928 0.270107 0.252507 NA
Which is not, I presume, exactly what you expected to get, but at least it runs.
One thing I often overlook is the fact that, as things stand now, you can't
define a function twice; to be more explicit: suppose you have a file
func.inp, which contains a function, and a file main.inp, which calls the
function via an "include" command. The first time you run "main"
everything
should work. The next time, you get an error because the second call to
"include" makes gretl think you're defining a function whose name is
already
taken. The solution I use is to put a "clear" statement before the function
definition; example:
(contents of func.inp)
function foo clear
function foo(scalar x1, series x2)
...
so I can include func.inp as many times as I want.
Hope this helps,
--
Riccardo "Jack" Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona
12:20 p.m.
? print y1
print y1
Varname: y1
Full data range: 1 - 20 (n = 20)
Current sample: 1 - 20
0.800000 0.820000 0.728000 0.701200 0.644480 0.608392 0.565597
0.530435 0.494972 0.463206 0.432769 0.404711 0.378269 0.353663
0.330599 0.309071 0.288928 0.270107 0.252507 NA
Which is not, I presume, exactly what you expected to get, but at least it
runs.
Oh yes, this is what I wanted. I only have to plot it to obtain the
theoretical correlogram.
The problem seems to be with the Spanish translation, since your message made
me try it with the "gretl --english" command, and I obtained the same as
output as you, but running the script in Spanish, I have:
----
gretl versión 1.5.0
Sesión actual: 2006/02/08 13:12
? nulldata 20
rango de observaciones: 1-20
# function definition
? function corrAR2 (scalar phi1, scalar phi2)
? series y = 1.0
? scalar rho1 = phi1 / (1-phi2)
? genr y[2] = rho1
? genr y = phi1*y(-1)+phi2*y(-2)
? genr y2 = y(1)
? return series y2
? end function
# function call
? (y1) = corrAR2 0.4 0.5
? series y = 1.0
Generada vector y (ID 4)
? scalar rho1 = phi1 / (1-phi2)
Generada escalar rho1 (ID 5) = 0
? genr y[2] = rho1
Modificado vector y (ID 4)
? genr y = phi1*y(-1)+phi2*y(-2)
Reemplazado vector y (ID 4)
? genr y2 = y(1)
Generada vector y2 (ID 6)
? print y1
Nombre de variable: y1
Rango de datos completo: 1 - 20 (n = 20)
Muestra actual: 1 - 20
0,00000
--------
--
Ignacio Díaz-Emparanza
Dpto. de Economía Aplicada III (Econometría y Estadística)
UPV-EHU
12:55 p.m.
On Wed, February 8, 2006 13:20, Ignacio Díaz-Emparanza wrote:
> ? print y1
> print y1
>
> Varname: y1
> Full data range: 1 - 20 (n = 20)
> Current sample: 1 - 20
>
> 0.800000 0.820000 0.728000 0.701200 0.644480 0.608392 0.565597
> 0.530435 0.494972 0.463206 0.432769 0.404711 0.378269 0.353663
> 0.330599 0.309071 0.288928 0.270107 0.252507 NA
>
> Which is not, I presume, exactly what you expected to get, but at least it
> runs.
Oh yes, this is what I wanted. I only have to plot it to obtain the
theoretical correlogram.
I was just referring to the fact that I'd have left the 1 at the beginning and
I don't like the NA at the end, but that's a matter of taste.
The problem seems to be with the Spanish translation
The same with Italian:
gretl versione 1.5.0
Copyright Ramu Ramanathan e Allin Cottrell
Questo è software libero SENZA ALCUNA GARANZIA
Sessione corrente: 2006/02/08 13:55
"help" mostra una lista di comandi
run Ignacio.inp
/home/jack/Prove/Ignacio.inp aperto con successo
? nulldata 20
Periodicità: 1, oss. max.: 20,
intervallo: 1-20
? function corrAR2 (scalar phi1, scalar phi2)
? (y1) = corrAR2 0.4 0.5
? series y = 1.0
Generato il vettore y (ID 4)
? scalar rho1 = phi1 / (1-phi2)
Generato lo scalare rho1 (ID 5) = 0
? genr y[2] = rho1
Modificato il vettore y (ID 4)
? genr y = phi1*y(-1)+phi2*y(-2)
Sostituito il vettore y (ID 4)
? genr y2 = y(1)
Generato il vettore y2 (ID 6)
? print y1
print y1
Nome variabile: y1
Campione completo dei dati: 1 - 20 (n = 20)
Campione attuale: 1 - 20
0,00000
--
Riccardo "Jack" Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona
2:22 p.m.
On Wed, 8 Feb 2006, Riccardo Jack Lucchetti wrote:
One thing I often overlook is the fact that, as things stand now,
you can't define a function twice; to be more explicit: suppose
you have a file func.inp, which contains a function, and a file
main.inp, which calls the function via an "include" command. The
first time you run "main" everything should work. The next time,
you get an error because the second call to "include" makes gretl
think you're defining a function whose name is already taken. The
solution I use is to put a "clear" statement before the function
definition...
I wonder if we ought to get rid of that annoyance -- that is, to
permit the redefinition of functions without error.
Allin.
2:19 p.m.
On Wed, 8 Feb 2006, Ignacio Díaz-Emparanza wrote:
I wrote the following script, based on the information of the
manual, but the function does not work:
nulldata 20
# function definition
function corrAR2 (scalar phi1, scalar phi2)
series y = 1.0
scalar rho1 = phi1 / (1-phi2)
genr y[2] = rho1
genr y = phi1*y(-1)+phi2*y(-2)
genr y2 = y(1)
return series y2
end function
# function call
(y1) = corrAR2 0.4 0.5
Which is the problem?
Sorry, this was a numeric locale bug. The parameters to your
function were being read as 0 and 0, when running in a locale that
uses ',' as decimal separator. I have now fixed this in CVS.
A workaround is to define the parameters as scalar variables:
# function call
scalar arg1 = 0.4
scalar arg2 = 0.5
(y1) = corrAR2 arg1 arg2
But this workaround will not be necessary with the CVS fix.
Allin.
3:07 p.m.
On Wed, February 8, 2006 15:19, Allin Cottrell wrote:
Sorry, this was a numeric locale bug. The parameters to your
function were being read as 0 and 0, when running in a locale that
uses ',' as decimal separator. I have now fixed this in CVS.
A workaround is to define the parameters as scalar variables:
I hate to play "I told you so", but... woe on those who like locale-dependent
decimal separators!
--
Riccardo "Jack" Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona
6916
days inactive
6916
days old
6 comments
3 participants
participants (3)
-
Allin Cottrell -
Ignacio Díaz-Emparanza -
Riccardo Jack Lucchetti