El Monday 05 January 2009 21:16:44 alisson rocha escribió:
hi people,
Before my question, a little explain:
I have a time series (gasoline sales) from january:2000 to december:2007
and i´m adjusting a sarima model to make predictions. Well, i´m running a
sarima model with parameters (1,1,0)x(1,1,0). After running the sarima
model, by default, the prediction is made with 95% interval confidence
(represented by the green bars). My question is:
In gretl, can i make a prediction with another interval confidence (90% for
example)? I post the pic at
http://br.geocities.com/asilva_rocha/question_gretl.png to make the
question more clear.
Thanks all,
Yes, of course, but it is not as easy as point and click.
From the graphical interface you can only make a prediction with a
95%
interval confidence. If you use the gretl command-line console, or better, a
script, you can make interval forecasts with the command 'fcast'.
You can use something as the following script:
<script>
open /opt/gretl/share/gretl/data/data9-9.gdt
arima 1 1 0 ; 1 1 0 ; nocars
addobs 4
fcast --out-of-sample fnocars
series fcerr = $fcerr
scalar conf=0.95
series fnocars_inf = fnocars-critical(N, (1-conf)/2)*fcerr
series fnocars_sup = fnocars+critical(N, (1-conf)/2)*fcerr
</script>
'fnocars' contains the predicted values for nocars, 'fnocars_inf' is the
series with the inferior limits of the confidence intervals and 'fnocars_sup'
contains the superior limits.
You can change "conf" to 0.90 for obtaining the 90% confidence intervals.
--
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.et.bs.ehu.es