s-news
[Top] [All Lists]

regarima models in S+

To: "s-news@lists.biostat.wustl.edu" <s-news@lists.biostat.wustl.edu>
Subject: regarima models in S+
From: Liv Belsby <Liv.Belsby@aktivkapital.com>
Date: Fri, 9 Jan 2009 17:03:57 +0100
Accept-language: nb-NO
Acceptlanguage: nb-NO
Thread-index: Aclyc+CiXtqfl9XVS6GeX7ECpmXM4A==
Thread-topic: regarima models in S+

Hi,

I have estimated a RegArima model using the s+ function arima.mle, and selected a RegAR(3) model.  I try to figure out how I can reproduce the fitted

values (these coincides with   arima.filt(model)$pred, i.e., how the fitted value  of time t, say Y(t), is computed. The model is formulated

 

Y(t)=b0+b1*x1(t)+ …+bp*X(t)+a1*Z(t-1)+a2*Z(t-2)+a3*Z(t-3),

 

where the Z(t) is the difference between the observed Y(t) and the “prediction from the regression part”  of time t. All the coefficients are estimated by

the arima.mle function. However, I’m not sure how the Y(t) is computed.

 

I would appreciate if somebody could help.

 

 

Liv Belsby       
Portfolio Analyst
liv.belsby@aktivkapital.com

Aktiv Kapital
Innspurten 9, P.O. Box 6426 Etterstad, 0605 Oslo, Norway
Tel +47 22 91 59 52, Mob +47 95 96 11 29, Fax +47 22 91 57 05
www.aktivkapital.com

 

 

<Prev in Thread] Current Thread [Next in Thread>
  • regarima models in S+, Liv Belsby <=