R is Cool

R is an open-source "free software environment for statistical computing and graphics" (the FOSS answer to packages such as S, hence the name) which runs on most Unix platforms (including virtually any Linux distribution and MacOS X) and Windows.

Getting started

Download and install R on your favorite computer, and start experimenting with it. There's quite a bit of good contributed documentation, including several good tutorials, on the Manuals page.

• For folks on windows, Murdoch and Sutherland's R Tools is an invaluable resource.
• Statistical Computing with R: A tutorial - a lovely quick visual tutorial by Dong-Yun Kim of Illinois State University that PrathibhaMethil found for us in Winter '09
• Although it discusses version 2.1, so some of the information is out of date, An Introduction to R: Software for Statistical Modeling & Computing is an excellent in depth tutorial.
• Don't wait! Start using R packages as soon as possible, much, much sooner than you think you are ready to do so, as the package system is R's IDE...

Agile R: the packaging system

R's packaging system is the key to staying sane while writing your own R code. R follows agile software development practices such as keeping documentation and code together, and even using documentation as a potential source of executable unit tests. The packaging system provides gentle support to guide you towards those best practices.

Working with R packages (practice)

• Peter Rossi has wonderful resources to bootstrap software engineers into the R packaging system:
  • Start with: Some Useful R Pointers
  • Whether on windowz or not, Making R Packages Under Windows is a wonderfully detailed walk through of the package development tools. You'll also want the test package archive
• Another reason to use packages, is that you can then use R-Forge, R's answer to the Google code project. Collaborative Software Development Using R-Forge, Stefan Theußl and Achim Zeileis, R Journal, 1(1), 2009 explains how to use R-Forge, step by step.
• Creating R Packages, Using CRAN, R-Forge, and local R Archive Networks and Subversion (SVN) Repositories covers packaging from the perspective of collaborative development, discussing how the packaging system relates to CRAN, etc.

Why R's packaging systems works the way it does (theory)

• While not specific to the packaging system, John Chambers Facets of R tells the story of how R evolved the way it did, and therefore why the packaging system is the way it is. The comparison to Java and C++ is very valuable. Note that perl's objects are also functional, although perl is not exclusively object-oriented, unlike R. Btw, that difference is a strength, not a weakness, of both languages. We need different languages for different tasks and different communities.
• Richard Stallman explains that the 'free' in free software should be interpreted as the free in free speech, not free beer. That sort of freedom takes infrastructure. Collaborative Software Development Using R-Forge, Stefan Theußl and Achim Zeileis, R Journal, 1(1), 2009 (mentioned above) does a nice job explaining the role resources like R-Forge play in the lives of collaborative software communities.
• Literate programming refers to the practice of not just keeping documentation and code together, but exploiting that fact in various ways. R's packaging system does so to ensure, for example, that documentation is complete and is consistent with the corresponding code. Python and Perl offer modules (nose and Test::Inline, respectively, that take the process a step further and create tests from code samples in documentation.) See Literate programming for creating and maintaining packages, by Jonathan Rougier, R News, 5(1):35-39, May 2005 for the origins of literate programming in R. The current R CMD check offers much richer literate programming, as discussed in the materials in the previous section.

Extending R with C/C++/Fortran/Python

R has its own native programming interface. For more extensive uses, it can be can be programmed in a variety of languages, but my preference is PyThon (and I think you'll understand why, as we progress...), using RPy. You may want to install RPy and start playing with it. You could use C++ to program R, but I don't recommend it...

Datasets

To play with R, you'll need the datasets package, which is part of the core, but can be a bit hard to find for a beginner, so here's a direct link to the documentation: http://stat.ethz.ch/R-manual/R-patched/library/datasets/html/00Index.html

• Playing with data sets, by ShobhaPolepalli
• DataSets lists repositories of data sets useful for statistical pattern recognition.

R Resources

• The R Journal, and
  • R News, the journal predecessor
• Bay Area R Users
• The CSUEB statistics department is a hotbed of R activity
  • Powerpoint slides from an excellent seminar they hosted on R's package manager: Creating R Packages, Using CRAN, R-Forge, and local R Archive Networks and Subversion (SVN) Repositories, main seminar page.
• UnderstandingRPackaging - ShobhaPolepalli's journal understanding the R packaging.
• ScriptDocumentation.doc: Some notes on matrix operations in R
• The Ahat lab is into PatternRecognition, so we think the CRAN Task View: Machine Learning & Statistical Learning is pretty cool
• The R Mailing lists, especially r-help. The archives of the mailing lists (as well as much of the other material at the main R server) are searchable using several search engines. Links to those search facility are cataloged on the R project's search page