Date:         Wed, 20 Dec 2000 08:11:38 -0700
Reply-To:     "Stanley A. Gorodenski" <vvgsgor@DE.STATE.AZ.US>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
Comments:     To: tek <tek@SSB.NO>
From:         "Stanley A. Gorodenski" <vvgsgor@DE.STATE.AZ.US>
Subject:      Re: proc arima syntax
Content-Type: multipart/alternative;

I did not catch this thread of communication from the start, and so what I have
to offer may have already been addressed.

I don't know your level of knowledge of the arima method, but I, for one, would
not want to learn the syntax of sas arima based on SAS' description of the
arima method in their ETS manual.  Years ago I learned arima in a workshop that
had as a text book "Applied Time Series Analysis" by McCleary and Hay.
McCleary taught the workshop.  This is a very easy to read book and I highly
recommend it, but I believe it is out of print and is in libraries and used
book stores only.

My major experience with model building using the arima method is from using
the PACK program.  This program was given to all participants who had
participated in the workshop.  My primary purpose in writing this communication
is to point out an aspect of SAS' arima procedure that may not be up to your
expectations.   PACK will automatically plot the ACF's and PACF's of the
estimated model, which is useful for visual inspection as part of the model
building process.  However, SAS does not do this (at least with respect to
version 6.12) .  SAS will give you the ACF's and PACF's of the differences you
may specify in your model, but not of the estimated model (unless, of course,
the final model is composed only of differences, i.e., no autoregressive or
moving average terms).  To get this in SAS I have had to do an identify of the
residuals from the estimated model.  The following syntax illustrates what I
mean:

DATE=INTNX('QTR','01JAN60'D,_N_-1);
FORMAT DATE MONYY.;
IF Q1FOR20=0 & YRQTR > 1981;
PROC ARIMA; IDENTIFY VAR=Q1A20 NLAG=24; ESTIMATE Q=(1)(4)(6);
            FORECAST LEAD=60 ID=DATE INTERVAL=QUARTER OUT=ARIMOUT;
PROC ARIMA DATA=ARIMOUT; IDENTIFY VAR=RESIDUAL;

Stan



tek wrote:

> Kattamuri.Sarma@RESPONSEINSURANCE.COM skrev i meldingen ...
> >I don't know any internet sources for this. But SAS has atleast 3 books on
> >SAS/ETS which has
> >   syntax for ARIMA, and Time Series in general.
> >
> >  Hope this helps.
> >
> >   - Kattamuri Sarma
>
> Thanks for answering. Yes, I knew this. I haven't had a chance to look
> at these books yet. I have seen the documentation that comes with
> the installation of SAS 8.1,  but I think these documentations are not
> very good.
>
> Terje Karlsen

[text/html]