Re: [S] Comments on Concerns over New Versions of S-PLUS for Windows

Jens Oehlschlaegel (oehl@Psyres-Stuttgart.DE)
Mon, 25 May 1998 23:22:17 +0200 (MET DST)


On Mon, 25 May 1998, Steven P. Millard wrote:

> One thing to keep in mind is that like any good business, MathSoft
> wants to expand its market and grow. In order to do this, it has to
> create products that will reach broader markets. The people who use
> the command version of S-PLUS represent a small fraction of all people
> who use some sort of software to analyze data and produce graphs (I
> include people who use spreadsheets in this population).

Those rare people have been the core S+ users in the past, and they are no
longer: When I opened the 4.0 user manual I was shocked to find that
every thing was new and unknown but nothing is worth to invest learning
time: click here, click there. By experience, I know that this kind of
"knowledge" leads to a dead end. So may I savely drop the users manual in
the litter box? Will I find all important things in the programmers
manual? I really don't know.

May be I'm ignorant about market share, and underestimate the influence of
other people being fascinated by colored features and 3-D buttons, BUT:
when I (as a very experienced SPSS user, and thus reluctant to change)
came to know S+, I saw S+ in a really good starting position for becoming
*the* professional statistics software, because of it's convincing
professional concept. I'm not so convinced today.

I'm convinced of something like S+ because a language is much more
powerful than a gui, for professional use. For me it's a bad joke that in
the future a "standard version" of S+ does not contain the S language.

>From my little experience I find 4.0 rather irritating than helpful, which
I find even more annoying in the light of doubled fees since 4.0. I would
be happy to pay doubled fees for real improvements, by this I do not mean
adding a new method every 3 months but rather to rule out lots of
inconsistencies and unreliabilities. There are so many traps you can get
trapped in, so many little tasks you have to program on your own, (though
everyone should need them). For example it is really nice that that OOP
S+ automatically chooses predict.lm for an object of class lm. Well, if
we want a safe predict we should rather use predict.gam on lm objects. I
don't see the easy seeking gui user beeing convinced by such "features",
wether they are hidden behind a gui or not. There is still a lot to do on
the basics! Am I the only one waiting for an S+ compiler? Who would
voluntarily develop a statistical procedure in C if he could do it with an
S compiler? Wouldn't such intensified attraction of statistical
development be a strong investment into the future market share of S+?

I would pay quadrupled fees, if I could go with S+ alone and save the fee
for SPSS (or SAS). Up to now having SPSS gives me the save feeling, that
I'm equiped to handle large data bases. (I'm aware of SPSS's shortcomings
and it's 2 sets of syntax rules, +1 matrix language +1 macro language +1
script language not to speak about the new horrible output browser).

Finally:

I still prefer WinS+3.3 over 4.0, because even Release 3 is not stable on
my machine (and not fully compatible with my old code). I'm using 4.0 on
a pentium 133 with 64 MB RAM, Win 95, object manager disabled, and S+4.0
regularly crashes, typicall symptom: suddenly the RETURN key is ignored,
I can still type code at the command line, but not activate it. Of
course, 3.3 behaves quite well in the same environment.

Best regards

--
Jens Oehlschlaegel-Akiyoshi, a recently disappointed admirer of S

----------------------------------------------------------------------- This message was distributed by s-news@wubios.wustl.edu. To unsubscribe send e-mail to s-news-request@wubios.wustl.edu with the BODY of the message: unsubscribe s-news