🚀 go-pugleaf

> Has anyone out there had any experience using S-Plus (from Statistical
> Sciences) for MS-DOS?  Apparently version 2 is available now.
>
> Is this a good package to use for exploratory data analysis?  How
> are the graphics.  We already have SPSS and SAS for MS-DOS;  what will
> S-Plus permit me to do that these won't?  Comments?

A similar question was just answered on sci.math.stat and a great response
posted.  All I can add to what Bill said was that it does run under DOS and
runs pretty good.  You have to purchase a lot of extra stuff if you want to
add routines, but what is in S+ is usually enough.

-phil

		  o /             o /             o /             o /
  --Cut-here-------X---------------X---------------X---------------X----
		  o \             o \             o \             o \
From: wvenable@spam.ua.oz.au (Bill Venables)
Newsgroups: sci.math.stat
Subject: Re: opinions on S-PLUS? (longish)
Message-ID: <WVENABLE.92Jan6170503@spam.ua.oz.au>
Date: 6 Jan 92 06:35:03 GMT
References: <1992Jan5.164206.36742@kuhub.cc.ukans.edu>
Sender: wvenable@spam.ua.oz
Organization: Adelaide University.
Lines: 88
In-reply-to: geoff@kuhub.cc.ukans.edu's message of 5 Jan 92 22:42:06 GMT

In article <1992Jan5.164206.36742@kuhub.cc.ukans.edu>
geoff@kuhub.cc.ukans.edu writes:

>   I am considering purchasing the S-PLUS package from StatSci.  Does
>   anyone have any comments on it?  I realize that the 30-day trial offer
>   is available, but I doubt whether I would be able to take advantage of
>   that ...

The native article is S from AT&T, which often you buy as source.  Splus is
an enhanced version from StatSci that you buy for a specific platform as a
binary.  The enhancements are quite significant, but the native version
contains all the really essential stuff.  As it happens, we have both, (but
not on the same machine, of course!).

Also, it is important to note that it is not a "package" in the SAS / SPSS
/ MINITAB / ... sense.  It can best be described as a suite of facilities,
or tools, which together form a programming environment for data analysis
and graphics.  It is very Unix-like in look, feel and philosophy, and the
integration with the Unix operating system (or "interfacing" as they call
it) is very intimate.  It is also, like Unix, a true programming language.
The important thing to note is that it does require the user actually to
program the calculations, (although certainly not from scratch), rather
than simply to choose options from a menu.  However the programming tools
do allow you to focus on the statistical and data analytic aspects of the
problem rather than the computing minutiae.

Its advantages, as I see them, include

1. It is written by competent and experienced statisticians, and so for any
   data analysis task that comes up there is usually a number of safe,
   flexible and convenient tools available to handle it.

2. It is very strongly graphically oriented and it graphical facilities are
   top notch and well integrated with the other language facilities.

3. Proper attention has been paid to numerical reliablility.

4. With Splus you can easily hook in your own software written either in C
   or in Fortran and so the system is readily extensible even beyond the S
   programming language itself.  (This is usually possible with S, too, but
   not on all systems.)

5. It can be quite fun to use.  (We give our students introductory courses
   to several "packages" early in their fourth year and allow them to
   choose.  Invariably they pick Splus for their projects and just about
   all the other jobs we give them to do.  Most of them rapidly become more
   proficient than the staff.)

6. There is a large community of users and a growing collection of freely
   available software.

It's disadvantages include

1. It is written by people who are primarily statisticians rather than
   software engineers.  It's rather big and at times ungainly and in the
   wrong hands its use of memory and other computer resources can be
   profligate.  You can quickly lose all your friends who need to share
   your computer.

2. It can be a bit qirky.  The assignment operator is the eminently logical
   <- , where the rest of the world, practically, uses = or := .  Periods
   are allowed in indentifiers, but the underscore is not (in fact it is a
   synonym for <-).  The matrix multiply operator is the clumsy %*%.  The
   inverse matrix function is called "solve(..)".  It is a vehicle for some
   experimental developments in data analsyis software and new releases are
   often not fully upward compatible - a feature that infuriates some users.

3. It requires some commitment and regular use to remain fully comfortable
   with it.  It is more like playing a piano than like riding a bike.

4. It's pricey, but you get what you pay for.

5. It needs a fairly hefty system to run on.  Don't look for a mico version.

6. If what you really want is a WIMPy point-and-click system that purports
   actually to do your hard statistical thinking for you, then you are
   bound to be disappointed and upset.

>   Also, is there a newsgroup on S out there somewhere?

There is an active mailing list.  However as I get to it through a local
internet news group, I don't have the details handy.  Sorry.

[I am not associated with AT&T or StatSci.  However I do give courses in S
 for statisticians and data analysts.]
--
Bill Venables, Dept. of Statistics, | Email: venables@stats.adelaide.edu.au
Univ. of Adelaide, South Australia. | Tel: +61 8 228 5412  Fax: ...232 5670
		  o /             o /             o /             o /
  --Cut-here-------X---------------X---------------X---------------X----
		  o \             o \             o \             o \


--
     J. Philip Miller, Professor, Division of Biostatistics, Box 8067
	 Washington University Medical School, St. Louis MO 63110
	     phil@wubios.WUstl.edu - Internet  (314) 362-3617
uunet!wuarchive!wubios!phil - UUCP (314)362-2693(FAX)  C90562JM@WUVMD - bitnet