[R] The hidden costs of GPL software?

[Frank E Harrell Jr]

Mike Prager wrote:
> At 11/18/2004 07:01 AM Thursday, Thomas Schönhoff wrote:
> 
>> To sum up, what I am in need to is an extensive example based 
>> help-system, focused on how to do things in R. In parts this is 
>> already there, i.e. SimpleR from Verzani (contributed docs area) etc.
>>
>> Hopefully I can  contribute to this in future, since it is seems to me 
>> invaluable to learn R by going through example-based lessons (some are 
>> found in vignette() ).
>> These are much more comprehensible to me than those short reference 
>> like entries in the current help-system, mostly due to their very 
>> technical approach (same is to be said about the official GNU R 
>> manuals, especially "The R Language", which wasn't a great help for me 
>> when I took my first look at GNU R). In this context something like 
>> the GuideMaps of Vista come to my mind!
>>
>> But to be as clear as possible, I think GNU R is great and I 
>> appreciate all the efforts done by the R core team and associates!
>>
>> Nevertheless it seems to be valuable to re-think the help-system in R 
>> with respect to those who may have a good understanding in statistics, 
>> but lacking some basic experiences in how to introduce themselves to 
>> sophisticated world of R/S languages.
> 
> 
> (I posted similar material before, but it was moved to R-devel, and I 
> wanted to express a bit of it here.)
> 
> I have frequently felt, like Thomas, that what could make R easier to 
> use is not a GUI, but a help system more focused on tasks and examples, 
> rather than on functions and packages.  This has obvious and large costs 
> of development, and I am unlikely to contribute much myself, for reasons 
> of time and ability.  Yet, I mention it for the sake of this discussion.
> 
> Such a help system could be a tree (or key) structure in which through 
> making choices, the user's description of the desired task is gradually 
> narrowed.  At the end of each twig of the tree would be a list of 
> suggested functions for solving the problem, hyperlinked into the 
> existing help system (which in many ways is outstanding and has evolved 
> just as fast as R itself).  This could be coupled with the continued 
> expansion of the number of examples in the help system.
> 
> Now I must express appreciation for what exists already that helps in 
> this regard:  MASS (in its many editions), Introductory Statistics with 
> R, Simple R, and the other free documentation that so many authors have 
> generously provided.  Not to mention the superlative contribution of R 
> itself, and the work of the R development team.  It is beyond my 
> understanding how something so valuable and well thought out has been 
> created by people with so many other responsibilities.
> 
> Mike

...

I second all of that.  What you are describing Mike could be done with 
a community-maintained wiki, with easy to add hyperlinks to other sites. 
  Just think what a great value it would be to the statistical community 
to have an ever-growing set of examples with all code and output, taking 
a cue from the BUGS examples guides.  The content could be broken down 
by major areas (data import examples, data manipulation examples, many 
analysis topics, many graphics topics, etc.).  Ultimately the more 
elaborate case studies could be peer-reviewied (a la the Journal of 
Statistical Software) and updated.

Frank

-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University