Ed,
SAS is an old and venerable package whose roots date back to the
sixties. It is very complete and has excellent data management
capabilities. I would rate their documentation as being very good and
their graphics as being poor. For many years it was clearly the best
package and there are many statisticians who are familiar with it. I'm
sure that many employers require knowledge of SAS from their
statisticians.
I was a devoted SAS user until about 7 years ago when I attempted to
teach biostatistics to young physicians. My students were used to
computing with a point and click interface. SAS evolved at a time when
virtually all statisticians had programming experience with languages
like FORTRAN. If you are a skilled FORTRAN programmer you will find SAS
easy. My students, however, had never debugged a FORTRAN-like program
and their howls of anger when they first encounter error messages from
the SAS parser were not pretty. The problem with SAS is that you have
to write complete mini-programs (e.g. data steps) before the parser will
look at them. Often, the parser will not become really confused by your
errors until several lines after they were made. This makes its error
messages hard to decode by users who are only experienced with GUI
programs.
After my teaching disaster with SAS I looked around for a better
package. At that time, Stata was clearly the best package for my needs.
I believe that it still is. Stata is very complete, its documentation
is the best of any package I have looked at, and its graphics are
excellent. It is much easier to use than SAS. I have only just
received my copy of version 8, but I believe that its new GUI interface
will make it one of the most user-friendly of the comprehensive
statistical packages. Support from Stata Corp is phenomenal, as is the
number of Statalisters who are willing to help their fellow users. In
some ways, SAS has better data management capabilities, at least for
users with programming backgrounds. However, Stata's data management
capabilities are very good, and they are easy to use. Stata's
programming language makes it an excellent choice for statisticians who
wish to write their own statistical procedures, and there is a vast
quantity of user-written Stata programs that are readily available to
the Stata community (see, for example,. the Stata findit command).
Splus is another venerable language with roots in Bell Labs. It is well
liked by many professional statisticians, is very complete, and has very
good graphics. A weakness is that its designers wrote it using a
concept called bootstrapping. The idea was that they would write a
minimal core of procedures that could then be called by other procedures
that in turn could be called by even more general procedures. This
gives Splus an onion-like structure. A consequence of this for the
ordinary user is that, if you make a mistake, you get an error message
telling you that some procedure you have never heard of has taken
offense by something passed to it by another procedure that you also
have never heard of. What this means in practice is that when you make
an error, Splus will tell you that you have done something wrong, but
will not tell you what. Mending the error of you ways is up to you.
This makes Splus unacceptable to point and click users. (Splus has
grafted on a GUI interface in the past few years. I am not familiar
with it but have been told that it does not fully replace the old
command structure.)
In short, Stata, SAS and Splus are all excellent packages that can be
used effectively by professional statisticians. Stata, however, has the
additional advantage of being a first rate package for students, for
people who are not professional statisticians, and for users who do not
have strong programming backgrounds.
In my opinion, it is the best all around statistical package on the
market today.
Bill Dupont
-----Original Message-----
From: Edwin Earl [mailto:[email protected]]
Sent: Fri 1/24/2003 6:39 PM
To: [email protected]
Cc:
Subject: Re: st: On choosing a stats package...
I read with particular interest all replies to the query "On choosing a
stats package". I am curious to know how Stata does compare to Splus
and
SAS. Is true that SAS is the most required by employers?
Ed.
----- Original Message -----
From: "David Vaughan" <[email protected]>
To: <[email protected]>
Sent: Friday, January 24, 2003 4:38 PM
Subject: Re: st: On choosing a stats package...
> Thank you to the many people who responded to my query. That you were
> in favour of Stata was expected, but that the responses addressed my
> questions directly and thoughtfully answered the unspoken question: on
> the extent and nature of Stata support, both from the company and from
> its (evidently diverse) community. Not only the direct responses but
> the run of unrelated postings were part of the positive answer.
>
> I have a demo copy of JMP and am further exploring Stata through my
> daughter's university access. However, my conclusion is already there.
> I will be buying Stata and using Data Desk's EDA features when I feel
> that need.
>
> Thank you all for your assistance.
>
> regards
> David Vaughan
>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
