Welcome to Netrider ... Connecting Riders!

Interested in talking motorbikes with a terrific community of riders?
Signup (it's quick and free) to join the discussions and access the full suite of tools and information that Netrider has to offer.

Linux Gurus: Part 2

Discussion in 'The Pub' at netrider.net.au started by termis, Oct 16, 2009.

  1. I'm sure you've had enough of "which distro is for me" type questions, but I'm even wondering if I can get away with Linux in the first place. If any of you are linux + numbers nerd, this question is for you.

    I've been pretty keen about trying Debian, as the stability factor really appeals to me. Ubuntu sounds pretty good too, but the fact that it uses "unstable" kernels puts me off a bit -- I'd rather use proven stable, albeit not cutting edge, distros if possible.

    But the more important factor is that fact that I've come used to some reeeeally helpful Windows programs like SPSS and certain Excel plugins like Palisades DecisionTools software. These allow me to do some cool things like cluster analysis, linear programming, Decision Trees, Monte-Carlo simulations in a rather straightforward manner. I've tried looking for Linux freeware equivalents, but I'm not seeing much, other than OpenOffice being able to do some optimization stuff straight out of the package.

    ...And I just found out that I can download Windows 7 for US$30 with an email address that I have (an .edu address), so is the switch even worth it?
  2. For that kind of production stuff, I personally would be grabbing Windows 7 and sticking with it. Linux does lots of stuff well these days, but for just getting on and getting the job done with major commercial software I still choose Windows (hence the polyplatformal perversity).
  3. PS Can always set up a dual boot, of course...
  4. There is an open source alternative to SPSS, called PSPP (no joke), but it has very limited functionality. But you can use something like WINE to run Windows programs on a Linux platform. If it's mostly office functionality you want, Open Office is miles ahead of MS Office. It's faster and cleaner, except for one prob:

    The main reason that I haven't made the switch to Linux entirely (currently I run a dual boot) is that like most researchers I rely heavily on EndNote, and it's only the latest and greatest EndNote X3 that integrates with Open Office Writer. If you are running any other version of EndNote (which most of us are) you are tied to MS Word. And if you are tied to Word, you are tied to MS Office, so you may as well be tied to Windows.

    While I'm on this - never open up a Word doc loaded with EndNote fields in Open Office. Nothing will happen to your doc, but you can kiss the citations goodbye - it resets them all to plain text. I found this out the hard way :evil:
  5. I've tried PSPP but as whipster said, there's not much to it; basically just descriptives and basic hypothesis testing. You can use R though, which is free and works fine on linux, and does all the things you want with a bit of work on your part. An alternative to using EndNote would be to use LaTeX to write documents, then you can do bibliogs via BiBTeX, and again this is all fine under linux.
    As for the distro, I would use Debian in your position as you want stability; and since lenny went stable recently it's not that old either.
  6. Note that "Unstable" and "Stable" in Debian terminology has nothing to do with how much stuff will crash; it's more about how often changes are made to the software and then pushed down to the user. So, Debian "Stable" generally only releases updates when they really have to (security and major bugs). In Stable, you might get 3-4 package updates every month, Unstable would give you 10 or so, a day. A compromise is "Testing", which is where stuff goes before it ends up in Stable.

    This is great for a server where you basically just want it to be always up and only want to patch whats really broken. For a desktop, where you might want things that are a little newer (ie, newer browser, newer libraries, etc) to get the better functionality, then Ubuntu is probably a fine choice. I use 9.04 at work and am pretty happy with it.

    I use Debian Stable for my servers, and find Ubuntu to be a fairly solid desktop OS. There are annoyances with Ubuntu, mind, but it's not really Canonical's fault.

    I don't really have much of an opinion regarding your mathematical work, but if it were me, I'd be writing something in Python, using some open source/free/whatever python libraries to do the heavy lifting, to do the analysis and spit out the reports, rather than using a spreadsheet app & plugin. But then, I'm weird, and writing software seems to be what I do these days. I'm amazed people do anything more complex than track their budget in spreadsheet applications; they always seem so limiting to me.

    Probably the reason that you're not finding gui apps/plugins for linux that does your stuff is that it's there, just at a lower level than you're used to; a c or c++ library, some java frameworks, whatever.
  7. You must be a better man than I am! I've tried R a few times now, and each time I've wound up crying like a little girl and looking for my mum...

    I get so impressed when I look at the kind of things (especially complex figures) which can be done with R, it's just that the learning curve is so steep... I have to rely on a GUI like a nancy :cry:
  8. Ah, good stuff. Thanks folks.

    I'm desperately NOT trying to go the multi-boot route, as I'd rather have a single streamlined system rather than being a "fence sitter". Might not have an option, though.

    And I did take a took a look at PSPP, but wasn't impressed at all. It's got the basics -- regression, crosstabs, etc. but not much else -- no clustering, factor analysis, conjoint, MDS, etc...

    Gotcha -- thanks for clearing up the stable/unstable distinction.

    As for writing my own, yeah, small scripts for testing out a single process or so, but I can't imagine writing a full-blown program that would give you super-polished output like the palisades software would. I'm totally in awe of how awesome some of these number crunching software are.

    Plus, I'm definitely over that "hardcore" phase of my life. Even though spreadsheet plug-ins are slow for doing big simulations and like, they're just so damn convenient & pragmatic -- one can whip a presentable report in a relatively short span of time instead of battling over code.
  9. That got my attention because I have a .edu address too.
    Care to share how and where you can download it from for $30?

  10. As far as I understand, *.edu.au address will not work, though MS is supposedly releasing a Australian student price within the week as well.