Dubravko Dolic <dubro@DOLIC.DE> replied in high dudgeon: > Why learn SAS if You already know the best available Statistical > Environment: R. If You know a bit of R learning SAS is horrible. You > find Yourself sitting astonishing in front of the Monitor asking who > invented this #@*~ System called SAS. Tasks which R solves in seconds > or a few lines are nearly unsolvable in SAS. At least You find ways to > do it with SAS but there are from some other universe... > But as it seems: if You live in the UStates there are (not yet) ways > to get round monopolistic market structures which allow bad systems to > etstablish: means: Buy a book and find a job where they think they > need SAS.

I'm going to disagree with almost all of your points.

R is a nice package, and it is free. But it is simply not the same as SAS. It doesn't scale to large data sets like SAS does. It doesn't have the same capabilities. Even the package specifically designed to provide some PROC MIXED capabilities doesn't do anywhere near what PROC MIXED does, simply because PROC MIXED does so much. Coding your own solution in R (and then having to do all the testing, border-case checking, debuggung, etc.) when a working solution is only five or six lines of SAS code is counter-productive at best.

If you know R and only use programming languages which have a functional form, then yes, SAS can be confusing. As would be the case for C or Perl or Python or VB. But that doesn't mean the problem is the language. It means the problem is the programmer! In the same way, people who use SAS and are used to procedural languages like PL/1 and Pascal and Fortran find R to be mystifying and arcane and weird. Similarly, if you only know OOP and, say, Eiffel, then both R and SAS are likely to be confusing. I know from your posts that you are struggling with SAS, but that doesn't mean it isn't a fast, efficient, powerful system.

You say 'tasks which R solves in seconds or a few lines are nearly unsolvable in SAS.' I don't find that to be true. Perhaps that is because I've used SAS for long enough, and I used S-Plus before R came out so I was prepared for R. There are features of R that are really nifty, and I like the way that new statistical tools pop up in CRAN. But SAS has tons of nifty tools as well, and new statistical tools, and user-written code too. I find that a language like R or SAS or Perl or APL has a fundamental underlying set of systemata that you must grok in order to be able to fully use the language, otherwise you keep wondering why it isn't as good as what you used before. Your comment that the SAS methodology seems to be 'from another universe' suggests that you feel comfortable with the R Way but not with the SAS Way. That doesn't make one better than the other, except for you.

You seem to think that SAS has some manner of monopoly that keeps it in control of something or other. Check out any American university's stat or math department, and you'll see that simply isn't the case. SAS is simply one tool among many. It just does more things than most statistical tools. (Note that R and S-Plus are set up to use SAS data sets in lieu of their own database management systems - SAS didn't choose that.) The reason that SAS is used in industry jobs is because of its scalability to industrial-sized tasks. R doesn't have that - yet. S-Plus looks like it never will, despite the hopes of every one of its users. There are plenty of other alternatives out there, too.

So if you just don't like the look-and-feel of SAS, that's okay. Feel free to work in R. It's a nice program. I currently have as a side project the writing of some C code to speed up some chunks of R which run too slowly for data sets which are 3 orders of magnitude smaller than data sets I am currently running through SAS code to do similar things (it is for someone who doesn't want to learn SAS) so I know what R can and cannot do for me. I just don't expect R to scale like SAS, or to do all the things that SAS can. And I don't expect one of these packages to have coding style or structure like the other.

David -- David Cassell, CSC Cassell.David@epa.gov Senior computing specialist mathematical statistician