|Date: ||Wed, 8 Dec 2004 16:43:47 -0500|
|Sender: ||"SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>|
|From: ||Art Kendall <Art@DRKENDALL.ORG>|
|Organization: ||Social Research Consultants|
|Subject: ||Re: SAS faster than SPSS?|
|Content-Type: ||text/plain; charset=ISO-8859-1; format=flowed|
It is some years since I did any benchmarks and those were on IBM
mainframes. Running a procedure to select particular cases and save a
subset of the variables resulted in a bill about twice as large for SAS
as that for SPSS. Cost was a composite of cpu time, memory, and wall
time. At least at that time, without user intervention SPSS was smart
enough to know that if there was only one procedure it did not create
the scratch files. This was very useful in dealing with things like
Things will vary from procedure to procedure. But we used to figure
about 15% more people time to code SAS vs SPSS for the same job. It
also took longer to do the quality assurance review because SAS syntax
is not as readable (transparent).
At an extreme, on one job I did just before I retired, I did a QA review
of a 800 to 900 line SAS run by writing about a 30 line SPSS run and
seeing that results were identical.
SAS does a few esoteric data handling things more easily than SPSS. I
never had occasion to use them in something like 200 doctoral
dissertations and 1000 investigations on behalf of the Congress I have
consulted on. SPSS is very much more user friendly for the data
transformations that are part of almost every job. RECODE, AUTORECODE, etc.
A lot depends on exactly what you are doing. But because the GUI is so
useful in generating the first draft of the syntax, because of the
reduced staff time to develop and rewrite the syntax as the procedures
develop, because of the ease of syntax maintenance in long term
projects, and because of the greater ease in quality assurance review,
SPSS is often the major part of the toolbox on large projects.
When comparing packages it is important to compare them as they are
today. I have seen invalid comparisons of many kinds of software that
were based on versions from different years. 1990 vs 1996 etc.
Human factors in use of the software are a major factor in total cost of
Social Research Consultants
University Park, MD USA
Benny Kong wrote:
>I have heard that SAS is faster than SPSS, especially for processing large
>datasets. Using reasonably powerful hardware, what factor of performance
>increase should we expect when going from SPSS to SAS?
>Thanks in advance.