Date:         Fri, 8 Mar 1996 15:48:57 GMT
Reply-To:     David Nichols <nichols@SPSS.COM>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         David Nichols <nichols@SPSS.COM>
Organization: SPSS, Inc.
Subject:      Re: Time series in SPSS

In article <4hoa78$sbl@kettle.magna.com.au>,
Hong Ooi <irsdept5@magna.com.au> wrote:
>A couple of questions:
>
>- In ARIMA, I sometimes get warning #16567:
>>Our tests have determined that the estimated model lies close to the
>>boundary of the invertibility region.  Although the moving average
>>parameters are probably correctly estimated, their standard errors and
>>covariances should be considered suspect.
>
>Does this warning also apply to any other estimated parameters in the
>model?
 
No, just to the MA terms. If there were apparent problems with AR
terms, a different message relating to stationarity problems would
have been printed.
 
>- Can anyone give me some pointers on transfer function modelling in
>SPSS? How do I model Y as a function of X, given both X and Y have a
>(univariate) ARIMA error structure?
 
ARIMA doesn't explicitly deal with error structures on the right hand
side of the equation (transfer function-noise models). You can use
piecewise methods to deal with this kind of problem, but they're not
simple. The prewhitening approach recommended by Box and Jenkins is
discussed also in Abraham and Ledolter's _Statistical Methods for
Forecasting_ and in the BMDP documentation for the 2T procedure.
 
--
-----------------------------------------------------------------------------
David Nichols             Senior Support Statistician              SPSS, Inc.
Phone: (312) 329-3684     Internet:  nichols@spss.com     Fax: (312) 329-3668
-----------------------------------------------------------------------------