[6279] in s-news-athena
Ease of Use
daemon@ATHENA.MIT.EDU (Fred Wolfe)
Fri Jan 27 10:26:24 1995
Date: Fri, 27 Jan 1995 08:40:26 -0500
From: "Fred Wolfe" <wolfe%twsuvm.bitnet@utcc.utoronto.ca>
To: s-news@utstat.toronto.edu
The debate on 'ease of use' sheds a little light on the problem. I have been
learning S-Plus over the last few months, and putting in a great deal of
time doing it. Although I'm a quite amateur statistician, I can program
quite easily in SAS and BMDP, and in a couple of computer languages. Even so,
C-Plus has been the most difficult for me to learn. I have read the manuals
a couple of times, and I try to use the on-line help. I would like to make
these comments:
The on-line help seems often (not always) to be for experts. If you know
80% of the answer, online help is useful; if you know 20% it is not generally
useful. Its not organized well enough. Sometime what you want to know is
there, but you don't know it or can't find it.
The trouble with the manuals is that they are not organized usefully (IMHO).
The S-Plus manuals could be much better if they were critiqued at their next
incarnation by a newbie. They should be useful of all levels of users. The
more one learns about S-Plus the more these sources become useful. The
issues is whether the learning curve is too steep.
For a newbie, Chambers and Hastie (The white book) and Becker, Chambers, and
Wilkes (The Blue Book) are too complex (although the white book is a jewel).
I would have given up on S-Plus if it were not for 3 things. I discovered
the Venables and Ripley book and Phil Spector's, An Introduction to S-Plus.
Those books took me step by step through S-Plus and into analyses, and they
had USEFUL examples, something I had not found before. For this newbie,
Spector's description on how to run logistic regress was enourmously useful,
and not in the on-line help or manuals in useful, nebie type, detail.
The next thing that helped was the S-Plus archives. Absolutely couldn't
have done without them. A plea: more simple questions, code, examples and
discussion on S-News. It has seemed to me that 'dumb' questions are looked
down on, recently. Being dumb about many aspects of S-Plus, I learn a lot
from them.
Problems: I am part of the 'Windows' universe. To use items on Statlib,
I have to recode them (Ihave done it many times, but it is not always fun or
easy. Many of the 'great' items on statlib require C (in a particular dialect)
or Fortran. To some of us who do not use these languages/compilers, we would
be willing to trade speed for the ability to run within S-Plus alone.
Yesterday we were given a glimpse at the next version of S and an ftp site.
Yet when I got there it was all in postscript.
With 24 Mb of windows memory (I can run SAS on 8), I run out of memory
sometimes. I can only print one copy of a graph at a time, and some large
file operations give me GPF errors.
Don't get me wrong, I love S-Plus. But it could be made easier. much easier.
Fred Wolfe
Arthritis Research Center
Fax: 316-262-0761
