Social Science Statistics Blog

My anecdotal experience has been something like this:

Biostats Types, SAS
Economist Types, SPSS
Sociology/Psychology Tyoes, JMP
Machine Learning Types, Matlab
Biology Types, Excel and specialized analytical tools like Spotfire and GeneSpring.
Business Types, Excel (though often other types depending on how they got to be Business types)
Physics Types, Fortran :-)

Obviously, none of this is hard and fast--R and S-PLUS are present in all of these disciplines to varying degrees and individuals may use any of the tools if it does . Statisticians use a mixture of all of these (though not Excel so much).

In my experience:

Economist: Excel or other spread sheets :)
Biostats/Bioinformatics: R + Bioconductor
Ecostats: R, SPSS, a little bit of SAS/Base
Econometrics: E-views and recently R

(I am a statistician)

Economists: Stata is currently the most popular (but also SAS, EViews, Gauss, LIMDEP, whatever). I think some packages like SPSS and TSP are no longer widely used. For techniques that are more cutting edge (or ad hoc), Matlab or R are used (or even C, Java, FORTRAN).
Psychologists: SPSS and JMP are most popular as far as I can tell

Interesting question. I'm a former engineer and manager and now an independent consultant who participates on a program evaluation mailing list. I see program evaluators using a range of tools, with perhaps SPSS, SAS, and R being more frequently mentioned.

I used to use XLISP-STAT and still miss its brushing and other dynamic plots. For simple stuff, I use J more than anything, and I'll refresh my memory on R when I need to do something more complex.

If J came up blank for anyone, especially with respect to statistics, see http://www.cs.ualberta.ca/~smillie/Jpage/jtsp.pdf.

Answer: Whatever package is best for serious empirical work, both research and production. That's R. If you are not using R now, you will be someday.

PS. I do not think that the "e-mail address to receive notification of new comments to this entry" works. At least, it never has for me.

It's an interesting question. As someone who straddles psychology and machine learning, I tend to use SPSS for some applications and Matlab or C/C++ for most. I've used R, but it was in a statistics class offered through the Government department; and though I found it useful, I don't use it much anymore simply because nobody else in my lab or field does, and it's nice to be able to share code occasionally.

I'm an economist and prefer R.

Some economists not wanting complexity use Stata or Eviews. Other economists will use SAS, Matlab or GAUSS.

Interesting comparison here:

Keeling, Kellie B. & Pavur, Robert J. (2007). A comparative study of the reliability of nine statistical software packages. Computational Statistics & Data Analysis, 51, 3811–3831.