Date:         Sun, 3 Jul 2005 11:18:41 -0700
Reply-To:     Yiyu <shenyiyu@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Yiyu <shenyiyu@GMAIL.COM>
Organization: http://groups.google.com
Subject:      Re: What model I should choose, ARCH or ARIMA?
Comments: To: sas-l@uga.edu
In-Reply-To:  <200507031635.j63GZRnx019700@listserv.cc.uga.edu>
Content-Type: text/plain; charset="iso-8859-1"

ARCH and ARIMA deals with different things. ARCH means the variance of
epsilon is autoregressive, and ARIMA means  Rt is autoregressive and
has moving average. So you can see ARIMA deals with the first term at
the right of your model and ARCH deals with the second term at the
right of your model. You should be able to write an ARIMA model with
ARCH noise. Seems to me  ETS' proc autoreg ca do both, for example, to
fit a simple GARCH(1,1) model

  proc autoreg data = return ;
    /* Estimate GARCH(1,1) with normally distributed residuals with
AUTOREG*/
      model y =  x / garch = ( q=1,p=1 ) ;
   run ;
   quit ;