SAS
* better manuals.
* tech support for most universities contracted into the price, thus
for researchers.
* batch orientation. if you have to handle data sets that are as large
as your memory, SAS generally does it better. It seems to be an
"n-pass design." Years ago, when memory was expensive, I could not
use
S/R even for simple problems. Just a few simple operations, and I was
disk thrashing.
* all sorts of corporate-oriented data base and ready-to-go application
stuff, often not statistical in nature, at all.
R
* actually, I believe that perl---which can be used as R or SAS
backend---beats even weird SAS input statements in its flexibility.
though don't get me going on how crazy it is not to have in-code data
set embedding.
* a real programming language and a real graphics language.
* some stuff (e.g., built-in statistical procedures) are a bit overly
complex; other stuff is so beautifully simple and intuitive that it
borders on a miracle.
* interactive design.
both suffer from weird mysteries---magic incantations that gurus know,
and ordinary people cannot easily find.
and let me say---despite prof brian ripley's occasional grumpiness (
;-) ), he and the rest if the core R group have done absolutely amazing
things for the community, both building the program and in helping
support it on this forum. I wish some of the corporations or
universities that are using SAS would fund the R group a little, too.
regards,
/ivo
---
ivo welch
professor of finance and economics
brown / nber / yale