I am a graduating student,and want to get a job in quantitative finance.But I really need some advice on which statistical software is the best in this area,SAS,Stata,or R.Any suggestion?

I'd personally respond with learn as much as you can - be it languages or packages. The most flexible IMO is R. You can transfer skill to S+, in case your firm likes to pay to license it.R is free.R has literally thousands of add on packages on CRAN. Like perl.R is interactive and at the same time can be run on a grid in batch mode- easy to use and then leverageR is used by major players in the businessR has a great community will and quick to helpThe downsides are:R takes a bit of practice - as usual, more flexible means more ways to screw up.R *could* have lots of programmers, as anyone can get it. You need to own SAS to learn how to use it, more programmers may mean less payI personally never liked SAS. Supposedly it just chews up big data, but I never have 'big' data. Hardware is cheaper than annual licenses.Stata I can't honestly say I have ever seen anyone use, apart from some old econ profs I know.I'd say most 'quant' work will require copious amounts of C++ as welljust my 2 cents.

Of the three you mention, R is the only one I've ever heard used by quants. Personally I think that S (of which Splus and R are implementations) is a horribly designed language (objects but no implicit conversion operators, let alone ability to write your own - what a joke!), and that the GUIs of R and Splus are a joke; For creative quant work I'd recommend matlab - that's what I and the people around me use. The GUI/command line/source code/debugger integration is fantastic, the documentation is great, and the language has a much more natural feel, at least to me.On the other hand, a lot of smart people also like R and/or mathematica, so now while you've still got the time, I'd recommend you to become fluent in all three of matlab, R, Mathematica - they correspond to quite different ways of thinking about things so it's good to master all 3 (and C++ of course, but IMO only masochists and hardcore old-timers do _research_ in C++). Using the search function on these forums will give you plenty of threads discussing their comparative merits.Forget about SAS, SPSS, Stata, Eviews, etc.If you've got spare time after mastering the above, learn python and ocaml.

Finalguy, at what level are you graduating ?

S+/R is great statistical softwares as long as you know what you are doing. For quant works, I would recommend them as it quicker and easier than programming languagues.

I would say either R or Matlab. My preference would be R. The R GUI is not particularly wonderful, but that is really the whole point of R - its is a statistical language/environment as opposed to a statistical "recipes" package.

QuoteIf you've got spare time after mastering the above, learn python and ocaml. I would add Haskell and Ruby to that list.

Thanks for all the enthusiastic response.I probably will choose R.to DCFC,I will take my PhD defense soon.

QuoteOriginally posted by: rwinstonI would say either R or Matlab. My preference would be R. The R GUI is not particularly wonderful, but that is really the whole point of R - its is a statistical language/environment as opposed to a statistical "recipes" package.I disagree emphatically with the false dichotomy you seem to imply. A source-code oriented enviroment (which as far as I can see is your opposite of a 'recipes' package) does NOT mean you have to live with a shit GUI. Matlab gives you both the power of a pretty decent language (at least as good as R in terms of language design), and a great, mature, powerful GUI. The amount of specialized statistics packages is admittedly smaller (though by no means nil), but I've never missed those so far.

It's true that there doesn't have to be a dichotomy, however I think this applies more to commercial projects rather than open-source (unless you're one of the big OSS projects). Open source project teams are usually small and disparate, and hence have limited resources. Time spent on the GUI is less time spent on core features. It's true that R's Win32 GUI is not great, however I think the OSX GUI is much better.

Look no further than KDE to see what can be done with open-source GUIs.

Yeah, but in KDE's case the project mainly _is_ the GUI.