5:37 p.m.
Dear Gretl Community,
Is there a way to circumvent the fact that we can not use the command
"dataset" inside a function? Please take a look in the following code:
<hansl>
function series TESTE (series Y, int tamanho)
dataset addobs tamanho
string nome = argname(Y)
print nome
gnuplot @nome Y_hat_$P_1_$Q --time-series --with-lines --output=display
end function
dataset clear
open fedstl.bin
data exbzus usphci
TESTE(exbzus,48)
</hansl>
Best regards,
Henrique Andrade*
*
6:21 p.m.
New subject: Information criteria for Quantile regression
Dear Gretl users,
Could anyone gist me on the information criteria in Quantile
regressions. Any goodness of fit criterion is also welcome. To be more specific,
information criteria for OLS are the coefficient of determination(R²: explanatory power of
estimated coefficients), Fisher statistics(for overall model) and student stats(for
estimated parameters). Information criteria for VARs are based on optimal lag
selection(AIC, BIC, HQC..etc), standard errors and so on. Are there other information
criteria, beside standard errors in Quantile regressions? Has anyone in here implemented
the "goodness of fit" process developed by Koenker and Machado (1999)?
Cheers
________________________________
From: Henrique Andrade <henrique.coelho(a)gmail.com>
To: Gretl Discussion List (users) <gretl-users(a)lists.wfu.edu>
Sent: Wednesday, April 18, 2012 7:37 PM
Subject: [Gretl-users] Add observations inside a function
Dear Gretl Community,
Is there a way to circumvent the fact that we can not use the command "dataset"
inside a function? Please take a look in the following code:
<hansl>
function series TESTE (series Y, int tamanho)
dataset addobs tamanho
string nome = argname(Y)
print nome
gnuplot @nome Y_hat_$P_1_$Q --time-series --with-lines --output=display
end function
dataset clear
open fedstl.bin
data exbzus usphci
TESTE(exbzus,48)
</hansl>
Best regards,
Henrique Andrade
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
6:55 p.m.
New subject: Information criteria for Quantile regression
On Wed, 18 Apr 2012, Anutechia Asongu wrote:
Are there other information criteria, beside standard errors
in Quantile regressions? Has anyone in here implemented the
"goodness of fit" process developed by Koenker and Machado
(1999)?
When you run a quantile regression in gretl you get AIC, BIC
and HQC, based on the loglikelihood as calculated by Koenker
and Machado.
Allin Cottrell
4:01 p.m.
On Wed, 18 Apr 2012, Henrique Andrade wrote:
Is there a way to circumvent the fact that we can not use the
command
"dataset" inside a function?
No. By design, functions are supposed to take the dataset as
given by the caller, and not mess with it.
Please take a look in the following code:
<hansl>
function series TESTE (series Y, int tamanho)
dataset addobs tamanho
string nome = argname(Y)
print nome
gnuplot @nome Y_hat_$P_1_$Q --time-series --with-lines --output=display
end function
dataset clear
open fedstl.bin
data exbzus usphci
TESTE(exbzus,48)
</hansl>
You have to do the "addobs" part outside of the function. But
note that this sort of function would likely be problematic
anyway. What happens if I do
TESTE(exbzus,48)
TESTE(usphci,48)
As written, I'd end up with 96 added observations: are there
plausible conditions under which that's what I'd actually
want? I don't know.
Allin Cottrell
4611
days inactive
4612
days old
3 comments
3 participants
participants (3)
-
Allin Cottrell -
Anutechia Asongu -
Henrique Andrade