R GUI ( was Re: [R] R Documentation(s)

[Peter Dalgaard BSA]

"Paul E. Johnson" <pauljohn at ukans.edu> writes:

> Peter Dalgaard BSA wrote:
> > 
> > Actually, I think there are valid reasons for people to want GUI
> > interfaces to statistics. One argument that I hear frequently is that
> > nonstatisticians tend to do other things than analyse data for
> > extended periods of time, and it becomes difficult to remember the
> > formalities from one time to the next. So they get scared of the
> > prospect of staring empty-minded at the R prompt. 
> > 
> > As long as you can turn the GUI *off* and use R as usual, I see no
> > reason not to have some form of GUI. 
> 
> I used SAS to produce some reports that I couldn't make R generate. 
> Perhaps R can, I couldn't see how. 

It usually can, but not without end-user programming...

> The reports from SAS PROC TABULATE
> summarize about 100 variables, showing the means & other summaries for
> subgroups to allow a side-by side comparison.  

Yes, this is the kind of thing SAS is really good for. In R you'd end
up with stuff like

data(airquality)
attach(airquality)
do.call("cbind",lapply(airquality, tapply, Month, mean, na.rm=T))

or (except for some field width trouble)

by(airquality, Month, function(d) sapply(d, mean, na.rm=T)) 

If people need this kind of functionality we should probably consider
including more of it. The real problem would seem to be more in the
interface specification than in the actual table generation. As usual,
contributions are welcome. On the other hand this kind of report might
be better handled by an actual database system - R easily gets in
trouble just by storing data frames with hundreds of variables.

> I had not used SAS for
> several editions, but in edition 8 they have a thing called "Analyst"
> where you can point and click on various anaysis/graph types and it
> generates a code template, which you can then grab, copy, edit, etc. 

This mode of operation would probably also be what we'd want in a
full-blown R GUI. The scratchpad notion also needs to come in
somewhere, the line-based interface has some shortcomings (we have it
in ESS, but then people would have to learn emacs).

> The SAS manuals are, well, almost impossible to understand.  Take a look
> at the writup on PROC TABULATE if you don't know what I mean.  But with
> analyst, they had a few templates for tables, and I could get the code
> for them, and get what I wanted after that.

I have to disagree on the SAS manuals there. Dauntingly verbose they
are, but unintelligible, not in my experience. And yes, examples are a
good thing. One of the problems we have with the current documentation
is that neither the demos nor the help page examples work really well
for getting users acquainted with the basic modes of operations.
Interactive tutorials may be the ticket, but they have to be written
first. I'm a little less sure that displaying R function calls are
quite enough for the new user (and for some things to work in a GUI,
we might have to include ugly and complicated constructions to avoid
having the GUI fall in the same pitfalls as users do.)

[snip]
> One of the most intimidating things about R is the seeming endlessness
> of it.  I've fiddled with it since version 0.64 or so, and I've
> contributed RPMs for the hdf5 data format in R, so I'm not completely
> dead weight (just mostly so).  But after all this time I don't quite
> have a mental image of its capabilities and boundaries.  A GUI helps
> people to see what can be done and how it might fit together.

Well, yes, but... One slightly arrogant way of saying it is "R is
a programming language, it's *supposed* to be endless". GUIs generally
work by creating a bounded universe, which can be mapped, and
therefore inevitably limit the user. That being said, there are quite
a few "neat tricks" and programming paradigms that don't really dawn
upon users until very late in their development. The use of the *apply
functions for one thing and the conventions for model objects, summary
and print methods for another.  


-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._