Date:         Tue, 15 Aug 2006 13:57:34 -0500
Reply-To:     "Weeks, Kyle" <kweeks@spss.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         "Weeks, Kyle" <kweeks@spss.com>
Subject:      FW:      Re: SPSS question
Content-Type: text/plain; charset="us-ascii"

Just a follow-up to one of Dan's points regarding SPSS and large datasets.  We have users that process terabytes of data with SPSS, including both very wide and very long files.  Processing time on procedures is linear in the number of cases in most procedures for the accumulation stage, but there are some procedures that require data in memory due to the nature of their algorithms.  The Data Editor, which is by nature interactive, is not well suited for very large numbers of cases. Thus, things like scrolling through a large dataset in the Data Editor would not be recommended.

Here's a simple experiment, with processor time results:

*1,000 cases, processor time: 0:00:00.01.
input program.
loop #i=1 to 1000.
do repeat x=var1 to var5.
compute x=rv.normal(100,10).
end repeat.
end case.
end loop.
end file.
end input program.
correlations variables=var1 to var5.

*10,000 cases, processor time: 0:00:00.16.
input program.
loop #i=1 to 10000.
do repeat x=var1 to var5.
compute x=rv.normal(100,10).
end repeat.
end case.
end loop.
end file.
end input program.
correlations variables=var1 to var5.

*100,000 cases, processor time: 0:00:00.78.
input program.
loop #i=1 to 100000.
do repeat x=var1 to var5.
compute x=rv.normal(100,10).
end repeat.
end case.
end loop.
end file.
end input program.
correlations variables=var1 to var5.

*1,000,000 cases, processor time: 0:00:07.84.
input program.
loop #i=1 to 1000000.
do repeat x=var1 to var5.
compute x=rv.normal(100,10).
end repeat.
end case.
end loop.
end file.
end input program.
correlations variables=var1 to var5.

Regards.

Kyle Weeks, Ph.D.
Director of Product Management, SPSS Product Line
Product Management
SPSS Inc.
kweeks@spss.com
www.spss.com
SPSS Inc. helps organizations turn data into insight through predictive analytics.


-----Original Message-----
From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of Dan Zetu
Sent: Tuesday, August 15, 2006 11:55 AM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: Re: SPSS question

Lisa:

You would have to admit that this is a broad question. As a user of both
SPSS and SAS, in general I personally find the two packages as offering
comparable capabilities, although there are also some differences. SPSS has
done a pretty good job in replicating most of SAS capabilities, but also has
added some that SAS does not have. Please be advised that my comments below
are based solely on personal experience and by no means they include the
entire universe of possible comparisons between the two.

1. SPSS cannot directly handle large data sets (at as low as 10000 cases,
its performance degrades). So if one has to work on a very large data set,
SAS is the number one choice

2. SPSS has a far superior GUI than SAS, the feature I find the most useful
is the capability to paste the syntax when using a menu command. This can
save one from writing a lot of syntax

3. In terms of statistical analysis, both packages offer quite similar
capabilities. There are a few offerings in SAS which do not have
correspondent in SPSS (as far as I know), such as PLS regression, superior
hierarchical clustering diagnosis tools, generalized additive models,
however SPSS has some other capabilities I have not found in SAS, such as
Optimal Scaling (or homogeneity analysis), which I found very useful in
certain instances.

So it really depends on the context of your friend's job description. For
example, if she is a chemist, SAS could be more useful as it has PLS
regression, which is heavily used in chemometrics. Same holds for some
direct marketing applications, and virtually all other instances in which
large data sets are used. Otherwise, there should not be any difference.

I am sure you will get even more complete responses here.

Dan

>From: Lisa Stickney <Lts1@ptd.net>
>Reply-To: Lisa Stickney <Lts1@ptd.net>
>To: SPSSX-L@LISTSERV.UGA.EDU
>Subject: SPSS question
>Date: Tue, 15 Aug 2006 12:37:11 -0400
>
>Hi All,
>
>     My apologies for this slightly off-topic post, but I've received an
>e-mail from a friend with a request -- to compare the functions of SAS &
>SPSS, but I don't know SAS very well.  I will post an excerpt from this
>message, and if you can help, please contact me off-list.  Thank you.
>
>I need to be able to make an argument that SPSS and SAS are comparable.
>This is for a friend whose green card is potentially at risk because the
>people at INS are thinking that she needed SAS to be qualified for her job
>here -- I don't think they have a clue what either program really is,
>they're just looking at a checklist and potentially deporting her on this
>technicality.   So, can you point me to any official or unofficial
>comparison of the two and/or do you have any insight to share?  In
>particular, feel free to be "devil's advocate", if you know of things that
>can be done in SAS that cannot be done in SPSS, etc.
>
>    Best,
>         Lisa
>
>P.S.  I've already directed her to UCLA's web site.
>
>
>Lisa T. Stickney
>Ph.D. Candidate
>The Fox School of Business
>      and Management
>Temple University