home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!SUVM.BITNET!SASMAINT
- Message-ID: <STAT-L%92121411534868@VM1.MCGILL.CA>
- Newsgroups: bit.listserv.stat-l
- Date: Mon, 14 Dec 1992 11:05:19 EST
- Sender: STATISTICAL CONSULTING <STAT-L@MCGILL1.BITNET>
- From: "Nelson R. Pardee" <SASMAINT@SUVM.BITNET>
- Subject: SAS vs SPSS
- Lines: 51
-
- One of the areas barely touched in the previous discussion is
- support. A battle I currently fight is the proliferation of STAT
- packages on campus, and not because I think that any one package does
- everything. When someone needs help, they come to our Computing
- Services organization. Simple arithmetic comparing the number of
- user support personnel to the number of packages we are expected to
- support says that we can support one package semi-well.
- Some faculty members endeavor to learn and know
- what they teach and hence can help their students. Sadly, many do
- not (and students come to us first, regardless of what they are
- told). There is need for some students (depending on interest/
- discipline/ advancement in learning) to learn a variety of packages.
- However, it's generally not productive for folks to struggle with
- unfamiliar software without being able to get good help.
-
- Another factor is that software costs are becomeing increasingly
- problematic. In particular, since it is constantly updated, it
- becomes a fixed cost in one's budget. SAS gets rapped because
- everything they sell (except JMP) requires a yearly fee for new
- releases, support, etc. But you get hit anyway with "one time
- purchase" software; when the new version comes out, the resulting
- clamor for it forces you to buy the next upgrade. We're getting
- eaten alive with this; and every piece of software added makes it
- worse. It's like the family budget; the fixed expenses eat up all of
- one's income.
-
- Finally, I'd like to put in my two cents for "accurate" packages. I
- keep hearing things (here and on the net) about a "nice" package with
- "wonderful" graphics and a "friendly" interface. Folks, these aren't
- cars, where the selection criteria is relatively straightforward. We
- wouldn't buy a car no matter the style and ergonomics unless it
- "drives" and we expect it to be reliable. Unfortunately, with a
- stats package we must take a great deal on faith about what goes on
- inside. Reliability is not usually about obviously failing; it's
- about correct answers. So what if the graphs are pretty? if the
- numbers seem nice. Are they correct? Can you tell? Sometimes. Ask
- somebody (yourself) the next time you choose/use your stats package:
- did you check its accuracy (or check with someone reliable who did)?
- Writing software is an enormously difficult task. Don't trust just
- anyone to do it right.
-
- The next time you want to add another Stat package, ask:
- -can I verify it gives correct answers?
- -will it be supported (there are costs if you support it yourself)
- -can we afford it
- -do I really need it
- There are reasons to proceed even if the answer is "no" to all the
- above; but at least, you'll have a better idea what you're getting into.
- --Nelson R. Pardee, Computing Services, Syracuse University --
- --120 Hinds Hall, Syracuse, NY 13244-1190 USA (315)443-1079 --
- --Bitnet: SASMAINT@SUVM Internet: SASMAINT@SUVM.ACS.SYR.EDU --
-