R

I am often asked to recommend a book for learning R.

Unfortunately, I have yet to encounter a book that I found to be "really" useful. R syntax never seemed intuitive to me, and hunting the documentation for names of functions etc, is ridiculously, pathetically, slow and dull. Hence, while I would like to say that the first day is the hardest, there are often long days of "hunting".

There are many fantastic things about R. It is by far the best statistics language out there. The data structures are very powerful, and in many ways, very elegant. It has fantastic packages, that cover a lot of econometrics, biometrics and statistics. And it is free: install anytime, anywhere. Buy a computer, install R. Easy. No separate price for each package etc.

Hence, if you want to venture into R-land, here is my note describing what I would recommend.

First, get TINN-R. Good clean simple text editor, that highlights the right stuff, and sends the code to R. There is a new IDE (Inference for R) that looks good, but I have not tried it. And there are other GUIs. TINN-R is where I would start.

http://sourceforge.net/projects/tinn-r

Second, read Lumley's notes to R. Very comprehensive, including a good treatment of graphics options in R. His treatment of graphics/plotting, is the best resource out there. http://faculty.washington.edu/tlumley/Rcourse/R-fundamentals.pdf

After 10 slides, you will probably be ready to quit. At that point, go over to amazon, and look up R textbooks. There are quite a few that are well regarded. Beg, borrow, steal ... or buy. The library may help. Dalgaard is the standard. While the books arrive, look at http://www.ats.ucla.edu/stat/r/

Do some tutorials from the site. Glance over other general documentation on R at http://cran.r-project.org/manuals.html

Be amazed by a few more of Lumley's slides. Awesome plots. Then glance at the more specific contributed documentation (generally longer and more detailed) at http://cran.r-project.org/other-docs.html

If you are coming from Matlab or Octave, then this document talks about moving from Octave/Matlab to R. It should be some what helpful, particularly if you know the syntax for something in Matlab. http://cran.r-project.org/doc/contrib/R-and-octave.txt

Octave implements Matlab syntax, and is free and open source. In general, it is reasonably regarded but I have never used it. My guess is that the majority of statisticians are on R rather than Octave, partly for historical reasons (R implements S, which was the statistician's language, while Matlab was always more the engineer's language). Partly because R organizes user conferences. And partly because R has a lot of the new packages. The momentum today is more behind R than Octave. But if you care, the general project is at http://www.gnu.org/software/octave/