In response to David Smith's message, a confession:
I too have been seduced by the GUI for my exploratory graphing needs. As a
(formerly) die-hard Unix hacker (since 7th Edition Unix on a PDP 11/70) my
original opinion was the same as Bert's: what a waste of time. But, speaking
of a big time-waster, I would regularly consome large quantities of time
trying to get complicated trellis graphs, with log-scaled axes, better
fonts, legends, annotations, home-grown panel functions, etc., JUST TO WORK
from the command line. After moving from Unix to S-PLUS on a honking NT box,
I finally decided to sit down for a day or so and just try to figure out
where they (Mathsoft's S-PLUS) were coming from with respect to graphics
GUIs. It was the best investment of my time I have made in a long, long
time.
I hardly ever go back to old-style graphics now: I have found that I can
get dynamite graphics in <10% of the time I used to spend getting the right
data into the right panel function at the right time, simultaneously
side-stepping S's quaint scoping rules. (Yes, I've thought about R, but the
GUI inteface for exploratory graphics in S-PLUS is a compelling reason for
staying put.)
So, if one of your job constraints is to get rapid, sophisticated analyses
with publication quality graphics out of your group's messy Excel
spreadsheets, the GUI will make the graphics part of that constraint
infinitely easier, and S-PLUS on NT will make the rest of the constraint
doable.
End of confession.
David O. Nelson
Lawrence Livermore National Laboratory
Box 808, L-441
Livermore CA 94551
Ph: 1.925.423.8898 Fax: 1.925.422.2282
> -----Original Message-----
> From: owner-s-news@wubios.wustl.edu
> [mailto:owner-s-news@wubios.wustl.edu]On Behalf Of David M Smith
> Sent: Wednesday, May 24, 2000 10:35 AM
> To: 's-news'
> Cc: 'Gunter, Bert'
> Subject: RE: [S] Editing labels in trellis graphs
>
>
>
> Speaking for my one opinion (and not for MathSoft), I actually
> like the GUI
> graphs. I find them useful for exploratory work, particularly
> when I get a
> new data set and want a quick graphical look at the variables. I'll also
> generally use the GUI for Trellis graphics, which I find a lot
> quicker than
> having to write a panel function for something beyond the basics,
> or having
> to remember exactly how to make the shingles for continuous data. That
> said, as an old S hacker at heart if I'm doing any kind of analysis at the
> command line I'll generally return to command-line graphics, both for the
> convenience of not switching windows and also because I generally want to
> keep a record of what I've done in the script file.
>
> I rarely convert between command-line and GUI graphs; it's
> generally simpler
> in the long run just to change the S code. (This is just a personal
> preference, though -- the conversion routines are actually pretty
> stable in
> S-PLUS 2000 Release 2, and I couldn't reproduce Pierre's
> particular problem
> in this release.) Here's a hint for Pierre -- you can add annotations,
> including text, to command-line graphs without converting to
> objects first,
> so one solution might be to draw the barchart without the label,
> and add it
> later using the annotation tool.
>
> As for MathSoft's decision to devoting time and effort to the GUI ... I
> wasn't around when that decision was made so I can't really comment there.
> My personal opinion is that it was a good one; the GUI is, IMO, a useful
> tool and better yet it's brought the S language to a range of
> statisticians
> and data analysts who otherwise would never have seen it, and as
> a long-time
> devotee of S that can only be a good thing.
>
> # David
>
> --
> David M Smith <dsmith@splus.mathsoft.com>
> S-PLUS Product Manager, MathSoft DAPD, Seattle WA
> Tel: +1 (206) 283 8802 x360
> Fax: +1 (206) 283 0347
-----------------------------------------------------------------------
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
|