[R] off-topic: better OS for statistical computing

[Scionforbai]

> My question is what is the best OS on PC (laptop) for statistical
> computing and why.

A free/open source *nix operating system is the best you can have
todate, for almost everything, noticeably stability, security,
scalability, networking and development, given that your hardware is
supported and the system is well administrated. Since you're talking
about a laptop, Linux has the best hardware support right now, and
should be your choice.

Linux is *very* flexible, stable, and gives you some good admin habit.
Is the best choice for number crunching, large databases and heavy
development. You have all kind of programming languages, compilers and
scientific softwares just a few commands away (pacman -S R, apt-get
install R, yum install R and so on) and they integrate perfectly
within your system. On the other side, you should have a certain
insight of this OS (and in general of computing) to reach the best
performances your hardware can give; most of user-friendly
distributions are as bloated as windows, well, maybe just a little
less, and if you don't know (and are not interested in learning) how
to tweak them, you don't have great advantages in terms of
performances and desktop experience.

My experience (mostly in university/research labs) shows that if you
really have to get into 'not basic' computing, involving for example
c/fortran/c++ development along with R/matlab/mathematica ('high
level' languages), or using for some reason clusters or distribuited
resources, then you really want to work natively on linux. You can
script and control everything, just as you want it to be, in a very
simple way, from the jobs schedule to the graphic interfaces; this is
for me the best feature of linux. For some the problem is that linux
(and in general unix software) tends not to be 'visual'. That's true,
and I consider it a major feature, when it comes to heavy tasks with
large datasets: grep, sed, awk and perl will always perform better
than excel or visual basic macros... and the advantage of linux on the
windows ports of these tools is that the whole system is built around
them, and they integrate perfectly in the powerfull shell.

If you just need to draw some picture to summarize data, and put them
in a WYSIWYG report/presentation, then the OS is not the point. If you
need to use  specific proprietary apps only available for win or mac
(word/excel/powerpoint/photoshop/flash/arcview/autocad...) then you're
locked in in that platform, but hey, that's not statistical
computing...

In my humble opinion: linux is by far best suited for development and
scientific computing, given the skills to admin it and the freedom
from platform-specific software. Mac os X 'just works' and is
definitely better and more attractive to me than windows.