R help - Nov 2007 - Packages - a great resource, but hard to find the right one

There have been several constructive responses to John Sorkin's  
comment, but none of them are fully satisfactory.  Of course, if you  
know the name of the function you are looking for, there are lots of  
ways to search — provided that everyone calls the function by a name  
that matches your search.  If you think there might be a function,  
but you don't know the name, then you have to be lucky in how you  
search.  R is a language and the suggestions so far seem to me like  
dictionary suggestions, whereas maybe what John is looking for is  
something more like a thesarus.

R packages are a strange collection, as befits a growing language.   
There are large packages, small packages, good packages (and not so  
good packages), personal mixtures of tools in packages, packages to  
accompany books, superceded packages, unusual packages, everything.   
Above all there are lots of packages.  As the software editor of the  
Journal of Statistical Software I suggested we should review R  
packages.  No one has shown any enthusiasm for this suggestion, but I  
think it would help.  Any volunteers?

Antony Unwin
Professor of Computer-Oriented Statistics and Data Analysis,
Mathematics Institute,
University of Augsburg,
Germany
	[[alternative HTML version deleted]]

Antony Unwin <unwin at math.uni-augsburg.de> [Thu, Nov 22, 2007 at
12:43:07PM CET]:
> There have been several constructive responses to John Sorkin's  
> comment, but none of them are fully satisfactory.  Of course, if you  
> know the name of the function you are looking for, there are lots of  
> ways to search ? provided that everyone calls the function by a name  
> that matches your search. 
I follow the suggestion to Google (mostly restricted by site:cran.r-project.org)
which gets me quite far.
> If you think there might be a function,  
> but you don't know the name, then you have to be lucky in how you  
> search.  R is a language and the suggestions so far seem to me like  
> dictionary suggestions, whereas maybe what John is looking for is  
> something more like a thesarus.
This is hard to do in a collaborative effort. One analogue is the
HOWTOs vs the man pages which I see in Linux. Some of the HOWTOs
are outstanding, the only problem they are facing is that they
tend to be out of date.
> 
> R packages are a strange collection, as befits a growing language.   
> There are large packages, small packages, good packages (and not so  
> good packages), personal mixtures of tools in packages, packages to  
> accompany books, superceded packages, unusual packages, everything.   
> Above all there are lots of packages.  As the software editor of the  
> Journal of Statistical Software I suggested we should review R  
> packages.  
You mean: prior to submission? 
> No one has shown any enthusiasm for this suggestion, but I  
> think it would help.  Any volunteers?
I am still putting some hope into the R Wiki. To my dismay it
is also package oriented, not method-oriented. I tend to think 
that there is a chance of controlled documentation if somebody
set out an infrastructure going beyond the current one. Anything
like a classification of methods. 

Thing is, I may like to volunteer, but not in the "here's a 
package for you to review by week 32" way. Rather in the way that
I search a package which fits my problem. One package lets me down
and I'd like to know other users and the maintainer about it.
The other one works black magic and I'd like to drop a raving 
review about it. This needs an infrastructure with a low barrier
to entry. A wiki is not the worst idea if the initial infrastructure
is geared at addressing problems rather than packages.

-- 
Johannes H?sing               There is something fascinating about science. 
                              One gets such wholesale returns of conjecture 
mailto:johannes at huesing.name  from such a trifling investment of fact.
http://derwisch.wikidot.com         (Mark Twain, "Life on the
Mississippi")

Johannes Hüsing wrote
> > Above all there are lots of packages.  As the software editor of the
> > Journal of Statistical Software I suggested we should review R
> > packages.
>
> You mean: prior to submission?
No.
> > No one has shown any enthusiasm for this suggestion, but I
> > think it would help.  Any volunteers?
>
> Thing is, I may like to volunteer, but not in the "here's a
> package for you to review by week 32" way. Rather in the way that
> I search a package which fits my problem.
That's what I was hoping for.
> One package lets me down
> and I'd like to know other users and the maintainer about it.
> The other one works black magic and I'd like to drop a raving
> review about it. This needs an infrastructure with a low barrier
> to entry. A wiki is not the worst idea if the initial infrastructure
> is geared at addressing problems rather than packages.
We should differentiate between rave reviews of features that just  
happened to be very useful to someone and reviews of a package as a  
whole.  Both have their place and at the moment we don't have either.

If you are willing to review an R package or aspects of R for JSS  
please let me know.

Antony Unwin
Professor of Computer-Oriented Statistics and Data Analysis,
Mathematics Institute,
University of Augsburg,
Germany
	[[alternative HTML version deleted]]

> Above all there are lots of packages.  As the software editor of the
> Journal of Statistical Software I suggested we should review R
> packages.  No one has shown any enthusiasm for this suggestion, but I
> think it would help.  Any volunteers?
There are two common types of review.  When reviewing a paper, you are
helping the author to make a better paper (and it's initiated by the
author). When reviewing a book, you are providing advise on whether
someone should make an expensive purchase (and it's initiated by an
third party).  Reviewing an R package seems somewhat in between.  How
would you deal with new version of an R package?  It seems like there
is the potential for reviews to become stale very quickly.

Another model to look at would be that of an encyclopedia, something
like the existing task views.  To me, it would be of more benefit if
JSS provided support, peer review, and regular review, for these.
Entries would be more of a survey, and could provide links to the
literature, much like a chapter of MASS.

Hadley


-- 
http://had.co.nz/

Antony Unwin <unwin at math.uni-augsburg.de> and Johannes 
Huesing <johannes at huesing.name> discussed ways to assist
navigation of the universe of R packages.

AU: 
>>> ...R is a language and the suggestions so far seem to  
>>> me like dictionary suggestions, whereas maybe what John
>>> is looking for is something more like a thesarus.
JH:
>> This is hard to do in a collaborative effort. One analogue
>> is the HOWTOs vs the man pages which I see in Linux. Some
>> of the HOWTOs are outstanding, the only problem they are
>> facing is that they tend to be out of date.
   ...
>> I am still putting some hope into the R Wiki. To my 
>> dismay it is also package oriented, not method-oriented. 
>> I tend to think that there is a chance of controlled 
>> documentation if somebody set out an infrastructure 
>> going beyond the current one. Anything like a classification
>> of methods.
What about Views? http://cran.r-project.org/src/contrib/Views/
Is this effort alive? Views allow easy downloading of packages 
grouped by methodology, with an accompanying overview.  But 
Views would seem to be at risk of going out of date.    
 
Or what about what an approach seen on many commercial sites?
What if CRAN package download pages had a mechanism for submitting 
reviews and for reading reviews of others?  I notice such reviews 
of books and software often mention competing products. For those
interfacing with CRAN via download.packages() and 
update.packages() commands, or via 'packages' menu items,
could these be amended to invite users to submit/read reviews?

Or this 'amazon.com' idea: if the community were not resistant
to a login mechanism, what if CRAN pages named 10 packages, 
related to the featured one by how frequently recent 
downloaders also downloaded them?

AU:
>>> As the software editor of the Journal of Statistical 
>>> Software I suggested we should review R packages.
>>> No one has shown any enthusiasm for this suggestion  
>>> but I think it would help.  Any volunteers?
JH:
>> Thing is, I may like to volunteer, but not in the 
>> "here's a package for you to review by week 32" way. 
>> Rather in the way that I search a package which fits
>> my problem.
AU: 
> That's what I was hoping for.
JH:
>> One package lets me down and I'd like to know other users 
>> and the maintainer about it. The other one works black magic
>> and I'd like to drop a raving review about it. This needs an
>> infrastructure with a low barrier to entry. A wiki is not 
>> the worst idea if the initial infrastructure is geared at
>> addressing problems rather than packages.
AU:
> We should differentiate between rave reviews of features 
> that just happened to be very useful to someone and reviews 
> of a package as a whole.  Both have their place and at the 
> moment we don't have either.
> If you are willing to review an R package or aspects of 
> R for JSS please let me know.
On the face of it, JSS reviews sound like a good idea.  But
is there something wrong with the www.jstatsoft.com site?
Today, at least, I cannot connect. Would a Google search on 
site:r-projects.org be likely to find a JSS review?  Would
the review be freely downloadable?  Would it too become out-
dated?  I'll contact you, Antony, about a review I may be 
qualified to write.

-John Thaden
Research Assistant Professor of Geriatrics
College of Medicine
University of Arkansas for Medical Sciences
Little Rock, Arkansas USA

Confidentiality Notice: This e-mail message, including a...{{dropped:8}}

R as a whole could benefit from systematic attention.
There may be scope for several special issues of JSS.

- R overview and philosophy

- R penetration and influence, in the statistics community,
   in machine learning, and in a variety of application areas.

- R as a vehicle for fostering communication between
    researchers in diverse areas.
    [A great thing about R-help, though nowadays this role
     is passing across to other lists such as R-sig-ME, is that
     it facilitates and even forces communication between
     application area specialists, and between those
     specialists and statistics professionals.  This may be
     temporary; we may see the R community fragment into
     diverse communities that focus on their own specialist
     interests? Scope for a sociological study, perhaps?)

- "Who is using R?", as reflected in published scientific literature.
     (I'd like to see a wiki or somesuch where authors
     are encouraged to give details of published analyses
     that have used R.)

- Where is R headed?  How will the shaping of its direction
     proceed?  Will it be a matter of step by step change and
     improvement, or is it (will it be) possible to lay out in
     advance an outline of the directions that its future
     development can be expected to take.

- Traps for new (and old) users.

- Books and papers on R.

- Then onto packages!  I guess what may be in order is
     something like the expansion of a task view into an
     extended paper.


John Maindonald             email: john.maindonald@anu.edu.au
phone : +61 2 (6125)3473    fax  : +61 2(6125)5549
Centre for Mathematics & Its Applications, Room 1194,
John Dedman Mathematical Sciences Building (Building 27)
Australian National University, Canberra ACT 0200.


On 23 Nov 2007, at 10:00 PM, r-help-request@r-project.org wrote:
> From: Antony Unwin <unwin@math.uni-augsburg.de>
> Date: 23 November 2007 7:50:52 PM
> To: r-help@stat.math.ethz.ch
> Subject: Re: [R] Packages - a great resource, but hard to find the  
> right one
>
>
> Johannes Hüsing wrote
>
>>> Above all there are lots of packages.  As the software editor of
the
>>> Journal of Statistical Software I suggested we should review R
>>> packages.
>>
>> You mean: prior to submission?
>
> No.
>
>>> No one has shown any enthusiasm for this suggestion, but I
>>> think it would help.  Any volunteers?
>>
>> Thing is, I may like to volunteer, but not in the "here's a
>> package for you to review by week 32" way. Rather in the way that
>> I search a package which fits my problem.
>
> That's what I was hoping for.
>
>> One package lets me down
>> and I'd like to know other users and the maintainer about it.
>> The other one works black magic and I'd like to drop a raving
>> review about it. This needs an infrastructure with a low barrier
>> to entry. A wiki is not the worst idea if the initial infrastructure
>> is geared at addressing problems rather than packages.
>
> We should differentiate between rave reviews of features that just
> happened to be very useful to someone and reviews of a package as a
> whole.  Both have their place and at the moment we don't have either.
>
> If you are willing to review an R package or aspects of R for JSS
> please let me know.
>
> Antony Unwin
> Professor of Computer-Oriented Statistics and Data Analysis,
> Mathematics Institute,
> University of Augsburg,
> Germany

	[[alternative HTML version deleted]]

Hi all,

I've been reading your suggestions and I'd like to share my thoghts,
which
might be common to people not so deeply involved in the development of R.

One of the advantages of R is that it's an open source software, meaning
that anyone can peruse the code for whatever purpose they might have. 

In R this is also true with contributed packages. However, for novices (like
me) R is terribly overwhelming and although I suspect there are excellent
professionals in their development, I really don't know how skilled the
person behind a certain package is, both in his discipline and in coding in
R.

Since it won't be short until I can understand complex R functions, in the
meantime I'd find a third-party review of the package useful. So I download
a package and know that at least someone else has taken a look at it and
vouches for its correct functioning or alerts of errors in certain points.
An alternative would be not to use R until I get to understand the code and
make sure it does what I expect, whereas I use another software to solve my
problems. In many cases this could mean setting R apart for good.

Also this is not a minor point if R is expected to spread at the corporate
level. Users need some level of assurance that what they use is reliable, if
not able to check the code in their own (having the code available is indeed
very useful, but not enough in all cases).

As for who should do the review, I wouldn't trust so much any user as I
would an organization such as the JSS. If the latter is not possible, the
first alternative is better than nothing.

Best regards,


-----
~~~~~~~~~~~~~~~~~~~~~~~~~~
Diego Mazzeo
Actuarial Science Student
Facultad de Ciencias Econ?micas
Universidad de Buenos Aires
Buenos Aires, Argentina
-- 
View this message in context:
http://www.nabble.com/Re%3A-Packages---a-great-resource%2C-but-hard-to-find-the-right-one-tf4855917.html#a13937354
Sent from the R help mailing list archive at Nabble.com.