Date: Thu, 5 Feb 2004 17:54:58 +0000
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
Subject: Re: SPSS vs. SAS, redux
Content-type: text/plain; charset=us-ascii
My knee jerk reaction to any statistical analysis delivered by a point and
click interface is to enquire about
1) the statistical qualifications of the person doing the analysis
2) the number of years they have used the software package
SPSS have made a deliberate attempt to make statistics point and click.
Virtually anyone can get a statistical analysis out of SPSS but do the
results mean anything?
I am not knocking SPSS, I just value the protection offered by a command
line interface and a modicum of "Technical know-how needed to operate".
Now a thought for the bean counters. Easy access point and click interfaces
in the wrong hands can result in poor analyses, produce bad decisions and
cost lots of money.
Katz To: SAS-L@LISTSERV.UGA.EDU
> Subject: SPSS vs. SAS, redux
Sent by: "SAS(r)
I thought now would be as good a time as any to dust off the periodic SPSS
comparison thread, and see what the current thinking is among my fellow SAS
users. (By the way, does SPSS have a list-serve comparable to SAS-L? Or
user groups like SUGI?) Naturally, I have an ulterior motive -- the bean
counters have once again recoiled in horror at the annual $A$ $hakedown,
and they want to do some comparison shopping. I've been appointed the SAS
defense counsel. The problem is, I know less than nothing about my
opponent. I haven't used SPSS in more than ten years; I've forgotten the
little that I knew, and I have no experience with the wealth of additions
they've made in the past decade. Here's the marketing pitch they hit me
* Integrated end-user, non-programmer interfaces
* Ease of learning for quick start
* Intuitive usability for productivity
* Complex operations built-in to by-pass programming
* Extensive help system includes tutorials, statistical
coaching and results interpretation
* English, Japanese, Chinese (traditional) and Korean
interfaces, help and documentation
* Comprehensive end-to-end functionality
* Open access to data in multiple formats (IBM DB2, ASCII,
Lotus, SAS, etc.)
* Wizard interface to data restructuring for analytical
* Automatic transformations (recoding, string, statistical,
* Comprehensive exploratory, analytical, and predictive
* Regression models, Advanced Models and Trends options for
* Deployable model-generated data results
* Analytical, interactive and presentation-ready graphics
* Pivot-able (slice-and-dice) and fully user-formatted
* Reports/graphics that can be copy/pasted, drag/dropped or
* Natural CRM and survey (people) data features
* Match/merge, transportation, aggregation of multiple tables
* Complete data field and value labelling
* Multiple missing data declaration, handling, analysis and
* Nominal, ordinal and scale-level measurement intelligence
* Customised, automated and integrated
* Customised operations via macros, commands and scripts
* Storage and re-execution of operations via commands, menus,
* Start-to-finish automated (batch) operations
* User-created default report and graphics formatting
* Speed, efficiency and resource-conservation
* Copy-free SQL queries
* Port-forwarding schemes for remote access
* Multi-threaded database access
* Multi-locale (national language) client and server sessions
My first observation is that SPSS and SAS modularize differently. What
SPSS calls Base would require SAS Base, Stat, and Connect. But a lot of
what they're saying, I can't tell if it's just marketing mumbo-jumbo, or if
there's real substance. For instance, when they say "automatic
transformations," do they just mean that they have a bunch of functions
that you can manipulate data with, like substring and mean and mod and log,
etc., or are they talking about a procedure like transreg that optimizes a
loss function, or do they mean something else?
What regression models does SPSS Base have, besides OLS? Logistic?
Censored via maximum likelihood? Censored via proportional hazards?
Response surface? GLM? Nonlinear? Does "trends" mean time series? Is
there another module with more models available?
Is "multiple missing data declaration, etc." anything comparable to PROC
What is "measurement intelligence?"
Is SPSS as programmable as SAS? When they say "macros, commands and
scripts" are they describing something like data steps and the SAS macro
language? (And is there anything in SPSS comparable to PROC IML?)
What are "copy-free SQL queries?" Does SPSS have the kind of full SQL
functionality of PROC SQL? Is "port-forwarding schemes for remote access"
like rsubmit in SAS/Connect? Is "multi-threaded database access" like
mpconnect; is it something SAS doesn't have?
One of the things you often hear is that SPSS is easier to use than SAS,
that it's more point-and-click oriented. Is SPSS Base comparable to SAS
Enterprise Guide in its point-and-click functionality? Is the SPSS input
wizard better than the SAS input wizard (I suppose it couldn't be any
worse!)? Can SPSS move data between hardware and software platforms as
easily as SAS?
Does SPSS have data manipulation and investigation procedures, such as PROC
TRANSPOSE, PROC CONTENTS, PROC RANK, PROC COMPARE, PROC EXPORT, PROC
FORMAT, PROC TRANTAB?
I welcome any answers you can give, and any references to detailed
comparisons. I'm also curious to know whether any of you have experience
in "mixed" environments, where some users have SAS and others have SPSS.
-- TMK --
This email has been scanned for Viruses and Spam. Please direct any queries
to Quest Information Security Support in Notes.
This email may be confidential, may be legally privileged, and is for the intended recipient only.
Unauthorised access, disclosure, copying, distribution, or reliance on any of it by anyone else is prohibited and may be a criminal offence.
Please delete if obtained in error and email confirmation to the sender.