Date: Fri, 6 Oct 2006 13:39:45 -0700
Reply-To: David L Cassell <davidlcassell@MSN.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: David L Cassell <davidlcassell@MSN.COM>
Subject: Re: SAS ARIMA Record Limitations?
Content-Type: text/plain; format=flowed
gorodenskis@MAIL.MARICOPA.GOV wrote back:
>Since no one replied by now I assume there won't be an answer to my
>questions. No problem. If anyone does respond sometime between now and
>the solar expansion projected to occur in a few billion years or so,
>please include my email address in the response. Since I have a tenuous
>connection with SAS (virtually none at the time) I need to set SAS-L to
>NOMAIL whenever I have no reason to get messages. I will do this by the
>end of the day.
> > _____________________________________________
> > From: Stanley Gorodenski - ASRX
> > Sent: Wednesday, October 04, 2006 10:23 AM
> > To: 'SAS-L@LISTSERV.UGA.EDU'
> > Subject: SAS ARIMA Record Limitations?
> > In my current job I am, yet, not in a position to be changing
> > procedures, but have a question about ARIMA and SAS. When I was using
> > the PACK program in the early 1980's on the mainframe, I was never
> > aware of any limitations on the number of records it could process.
> > There may have been a limitation but I was using small series, less
> > than 300 observations. When I had checked into pc Autobox about 6-10
> > years ago, it could only process at the most 600 records. My guess is
> > that the reason Autobox is/was limited to the number of cases that can
> > be processed is because the algorithms it uses are so time consuming.
> > Does SAS have any file size limitations for processing? If it doesn't
> > it probably means it is using a different parameter estimation method
> > than PACK(alias Autobox). Even if there is no size limitation, would
> > SAS be able to process up to 10k or 20k records in a reasonable time,
> > and, if not, will I always be faced with extremely long processing
> > times of large files no matter what software I use because of the
> > nature of the ARIMA process? In my current job we are using SPSS and
> > so I do not have access to SAS to directly answer my own question.
> > Stan
Sorry, Stan. I can't always be around. :-)
As I understand it, Autobox does not have that limitation anymore, and
has not for a few years. I don't know what Autobox's current limits are,
but they may only be restricted by the amount of RAM the software can
access on your machine. Go over to their site and ask.
As for SAS, the limitations tend to run based on the statistical
approach used to examine the time series. Tools like PROC ARIMA and
PROC AUTOREG can handle tens of thousands of records in a reasonable
amount of time. Probably a lot more than that. A tool like PROC UCM
that specifically works with a state vector will scale pretty much linearly
with the number of records and as the square of the number of entries
in the state vector.
David L. Cassell
3115 NW Norwood Pl.
Corvallis OR 97330
Search—Your way, your world, right now!