Hello, continuing the “Rolling estimate” topic…

 

I’m studying the command fcast and I see that it can be used in rolling mode just for OLS estimation models. Suppose I want to estimate a Logit model on a rolling sample of 48 observations over 1057:

 

      smpl full

      smpl 1 48

      loop for i = 1..1057

            logit Binary const ld_Close_1 –-p-values

            smpl +1 +1

      endloop

 

This gives me 2 * (1057 – 48) coefficients, which are “const” and “ld_Close_1”. Now I want to make a forecast using the same rolling window. I try this one:

 

smpl full

      smpl 1 48

      loop for i = 1..1057

            fcast i --static

            smpl +1 +1

      endloop

 

But it doesn’t work as I expected. Does anyone know how to produce an output like the GUI one, but using the rolling sample?

 

 

Am 27.05.2011 16:41, schrieb Oscar Soppelsa:

I want to express my congratulations to Professor A. Cottrell and Professor R. Lucchetti for the awesome gretl software. I'm studying the manual and it's impressive.

Indeed!

I have written this code in order to estimate an ARMA(1,1) model on I(0) time series using a moving window which uses only the last 20 values from the time series, "forgetting" gradually all the past information older than last 20 values.

 

The code (n = 1057):

 

      smpl 1 20

     loop for i = 1..1057

          arima 1 0 1 ; ld_Close

         smpl +1 +1

     endloop

 

Now I will try to create two new variables which are the AR(1) and MA(1) coefficients. Do you suggest me to use genr to create these variables? Does the command

 

genr c = $coeff

 

could allow me to create such a time series of coefficients?

smpl full
series AR = NA
series MA = NA

smpl 1 20
loop for i = 1..1057 --quiet
    catch arima 1 0 1 ; ld_close --quiet
    if !err
        AR[19+i] = $coeff[2]
        MA[19+i] = $coeff[3]
    endif
smpl +1 +1
endloop

smpl full

This code gives you the estimates you want if there are 20 observations included in the sample.

 

Thank you,

 

Oscar

-----Messaggio originale-----
Da: gretl-users-bounces@lists.wfu.edu [mailto:gretl-users-bounces@lists.wfu.edu] Per conto di Allin Cottrell
Inviato: Thursday, 26 May 2011 5:38 PM
A: Gretl list
Oggetto: Re: [Gretl-users] R: Rolling parameter estimate

 

On Thu, 26 May 2011, Oscar Soppelsa wrote:

 

> Thank Artur and Riccardo very much for your answer! I've two question now:

> 

> 1. could you explain me the syntax about the selection of the sample's

> dimension? What does this mean? ---> i=1982:1..1987:3

 

One standard form of loop in gretl uses the construction

 

loop i=1..10

 

which should be self-explanatory (but is also explained in the

manual). The variant above uses the convenience functionality of

representing integer observations numbers as dates (in this case,

quarterly dates).

 

> 2. After the rolling parameters estimation, how can I print on the monitor a

> vertical array whose elements are coefficients, std. errors, z, p-values

> etc.? I would need to export this array in Excel, for example.

 

Regression coefficients and standard errors are accessible in

matrix form as $coeff and $stderr, after estimating a model.

z-scores and p-values can easily be computed -- see the pvalue()

function. Matrices can be concatenated and reshaped at will, see

the chapter on matrices in the User's Guide. They can also be

printed in whatever form is appropriate - see the "printf"

command.

 

Allin Cottrell

_______________________________________________

Gretl-users mailing list

Gretl-users@lists.wfu.edu

http://lists.wfu.edu/mailman/listinfo/gretl-users

==============================
DISCLAIMER:
This e-mail is for the sole use of the intended recipient and any file transmitted with it may contain material that is confidential and privileged. If you are not the intended recipient of this e-mail, please do not read this e-mail and delete this message and any file attached from your system and then notify us immediately by reply e-mail or by telephone. You should not copy or use this message and any file attached for any purpose, disclose the contents of the same to any other person or forward them without express permission by us. Considering the means of transmission, we do not undertake any liability with respect to the secrecy and confidentiality of the information contained in this e-mail and in its attachments.
Il presente messaggio di posta elettronica e' ad esclusivo utilizzo del destinatario indicato in indirizzo e gli eventuali documenti allegati potrebbero avere carattere riservato. Qualora non foste il destinatario del presente messaggio Vi preghiamo non leggerlo, di cancellarlo dal Vostro sistema assieme ad ogni documento ad esso allegato e di volerci avvertire immediatamente tramite posta elettronica o telefonicamente. E' vietata la duplicazione o l'utilizzo per qualunque fine del presente messaggio e di ogni documento ad esso allegato cosi' come la relativa divulgazione, distribuzione o inoltro a terzi senza l'espressa autorizzazione del mittente. Il mittente, in ragione del mezzo di trasmissione utilizzato, non assume alcuna responsabilita' in merito alla segretezza e riservatezza delle informazioni contenute nel presente messaggio e nei relativi allegati.
==============================

_______________________________________________

Gretl-users mailing list

Gretl-users@lists.wfu.edu

http://lists.wfu.edu/mailman/listinfo/gretl-users