Hi,

I've recently converted to Linux and am checking out the options for statistical analysis.

(I know that most of you will have switched off by now!)

I'm well used to the package SPSS on Windows but is there any open source program that comes reasonably close in Linux? I'm not looking for perfection, but something that does crosstabs, Chi-square, Spearman's, ANOVA and simple regression would be great.

The only stipulation is that it would need to be GUI-based, rather than command-line run.

I've googled the subject, but I can only find programs that are a bit archaic and seem to be command-driven.

Does anyone know of any program (I know there's simple statistics in Open Office, but that's too basic) which would let me enter and analyse data in a straightforward way?

I've worked hard on getting ubuntu just the way I like it and a decent statistics package that means I'm not forever syntax-checking would complete my transition from Windows.

I look forward to hearing from the few of you who are still reading this!

Try R --> www.r-project.org. It is really a CLI application, but you can install the Rcmdr package for a GUI. This is a Tcl/Tk interface, but provides a "spreadsheet" type interface for entering datasets and menu-driven analysis. R will also import SPSS (among other) datasets. The Rcmdr package will also "log" everything you do which can be saved as a script for reuse. Yes, it is more than capable of the analyses you are interested in performing.

You can also use Gedit or gVim to write your scripts; both have R syntax highlighting. You can 'load' the script into an R session (it's as simple as source /path/to/your/script/script from the R prompt), do your editing in Gedit or gVim, and 'run' the script in R. If that doesn't make sense, let me know...after I read it, I'm not sure I understand what I was saying.

R is developed on a Debian machine and most of the developers are the guys who developed S.

Thanks for that. I'll check it out - and I will get back to you if I'm struggling to understand it!

Cheers.

I have not tried it, but I have heard of people using the windows version of Minitab under wine.

Sadly, I've only got a license for SPSS. I checked out wine compatibility and it's pretty dodgy. It starts the install procedure but hangs. I'd be happy to use Minitab (or SAS) but statistics software isn't cheap.

weibullguy - I've been looking at the r.project and it seems promising, especially as it has a ubuntu directory that's bang up to date. I'd like to find out more, but (forgive my newbieness) what is the Rmcdr package?

Is it specific to R, or is it something more general for building a GUI? I looked through the Feisty directory on R's CRAN but I couldn't find it anywhere.

Your advice would be appreciated - again!

Yes, Rcmdr is specific to R. After you install the base R package, start it from the command line. Then install Rcmdr (or any other package) That will install Rcmdr and all the other packages it needs. Exit RRestart R and load RcmdrHere is a quick (not high quality) screenshot of Rcmdr in action ---> http://webpages.charter.net/weibullguy/gfx/R.png
The available R packages that can be installed using the install.package command can be found at http://lib.stat.cmu.edu/R/CRAN/. Just follow the link called packages and scroll down the page to Rcmdr. Also save a copy of "The R Reference Index" found under the Manuals link. I have a short reference card somewhere. If I find it, I'll upload it and post a link.

Eventually you'll be using the second method I mentioned anyway (gedit or gVim to write the scripts).

I did a search to satisfy my own curiosity as I am also a fan of ANOVA. It was surprising to see that even Mac is better catered for than us. Anyway, I have come up with two possibilities which both include ANOVA and regression, and are GUI I think.

First one is http://www.statpages.org/miller/openstat/LinosOS.html

The second is http://www.gnu.org/software/dap/dap.html

By the way, if you want to be intimidated have a look at 'miller's' Background (oh dear, oh dear).

There is a third possibility in a book I bought recently

'Developing Statistical Software in Fortran 95'
by David R Lemmon & Joseph L Schafer (ISBN 0-387-23817-4)
I have been in Fortran since Fortran 1.5, and was on the Standards Committee which upgraded Fortran 77, but I find F95 heavy going. Consequently, I am going to stick to For77 or possibly For90.
The above book gives a website on page 12 where you could see source code from the book, and there is a lot of it.
http://methodology.psu.edu/fortranbook

Well, I've got R up and running from the terminal - but it's scaring the **** out of me.

I suppose I've been well spoilt in that I've only ever used SPSS in GUI mode. And it's a doddle. My professor at uni has told me about the days when he had to do everything through syntax only. I think he reckons today's SPSS users have it too easy.

Kintail, I'll definitely check out those links. (As for Miller's background, by the way, I look forward to reading it. Just for the record, when I was starting to study statistics, the worst book I ever found had the most inappropriate title - it was called Statistics Without Tears. I was weeping buckets by the end of the first chapter, for the author made the simplest procedures seem complicated. He almost scared me off even crosstabs.)

Weibullguy, As for R, I think it's got potential, but I'd really need to go GUI at the start before I can even begin to get my head around the syntax. I'm having problems there at the minute.

I found Rcmdr, as per your instructions, on the r-project page.

I did what you suggested and that got me into R, but the "install.packages..." line is only turning up syntax errors. I'm following the code exactly and running it from R.

I cut and pasted the code directly into R, so is it possible that I'm doing it wrong (remember, I'm still pretty much a newbie) and need to do some sort of edit on the line?

If not, is there any other way to get Rcmdr installed outside the R shell?

Really appreciate the help on this one guys. I didn't think a statistics thread would produce as much assistance as this one has.

Whao, I've just read Millar's background. Phew.

In the scientific area R is one of the killer apps of open source software. Compatibility with S-Plus there are lots of supporting examples and documentation. But it's certainly aimed at a math/stat audience. Even for people with programming experience I suspect it is not very intuitive.

I had a look on Freshmeat for alternatives. Searching on 'statistics' or 'statistical' is a bit too general but 'spss' gets a few hits.

Salstat looks interesting, although the last release was 2003, with disclaimers about lack of testing. Might be worth a look if you are stuck and it does what you want. It's written in python.

The Salstat author has a blog. It seems he is still working on the program, and writing a book to go with it.

Hehe. It takes a little while to get used to. You are installing software and, thus, you will need to run R as root when you install R packages. Otherwise, post your errors and maybe we can figure out what is wrong.

If you are using R for a university source, I'd guess you won't need any scripts that that in-depth. I might have some examples of regression, AOV, and DOE laying around. I do mostly survival analysis myself. I also found the short reference card I was looking for earlier.

The R short reference card is great.

But it's the first time I've seen a double-sided A3 sheet called a 'card'

I'm afraid I'm working in a Windows-only environment at the moment but when I get back to my Linux box at home I'll run R as root before trying to install.

If there are any problems, I'll post the output. Thanks for digging out the reference card. My university, unfortunately, is a totally Linux-free zone. There aren't even any Linux manuals on sale in the uni bookshop. Not one.

I'll check out the salstat pages too.

Cheers for now.

You can get R for Windows too. Then you can keep working on a consistent platform where ever you are. I'm surprised a University stats program isn't using primarily SAS though. Way tooooooooo pricey for the average student, but probably the most widely used in industry (in the US anyway).

Sorry for the delay in getting back to the thread, but I’ve been trying out the various suggestions so many of you have helpfully made.

So it’s time for an update. The best I’ve managed with R is to get it running from the terminal, but the Windows .exe actually works on Vista. A bit creaky, but it works, albeit with no facility for GUI data-entry.

Salstat also works in Windows, but it’s more than creaky. I tried various versions of the tarballs on Linux (some are incomplete, with a lot of dependencies required), while there is one bigger download that is self-contained. I got it unpacked but the executable won’t run, either from the folder or from the terminal. I think there is a .deb package (handy, as I use ubuntu), but the link’s not active.

My biggest success to date has been with OpenStat (the one that Mr Miller is responsible for) and, to be honest, it’s a crackingly good piece of software for GUI-users. It’s not too far away from earlier versions of SPSS in terms of what it can do. The Miller link above (see Kintail's post) is down but you can google to find the software elswhere.

The only thing is that I have to run it in Windows (it runs surprisingly well in Vista – the only unavailability is the Help menu, but it’s fairly intuitive for someone used to SPSS). Mr Miller is to be congratulated – the help file thing is just a Vista problem.

There is a Linux version of OpenStat but the download links are not working. So if anyone can find me an active link to the package for Linux I’d be very grateful. It’s a great program and I’d love to run it in Linux. When I find the time I’ll see if I can get it running under Wine. But if you find that active link, please let me know!!!

The drawback with most of the above is that you can’t import SPSS data directly and unfortunately, because my own version of SPSS just won’t work on Vista, I can’t access my files to re-save them in an appropriate format for import. I'm graduating next week and don't think I'll have the chance to convert all my datasets in uni.

Just a final word or two on R for weibullguy. The GUI in R looks promising, especially if I could get entering data. This, again, might be a Vista problem. But I’d like to have another go at getting it up and running in Linux, so here’s the output when I try to access Rcmdr as per your instructions.

After the code, running as root:

install.packages("Rcmdr", dependencies=TRUE)

R returns:

Error: syntax error, unexpected AND, expecting ',' in "install.packages(&"


As a footnote, it turns out that SAS was there all the time, but I was only ever trained in SPSS and just looked for it automatically.

By the way, I hope the thread has been useful for those who work with statistics but don't have the big bucks for the big name software.

If I can either get the R GUI running in Linux, or get a working Linux version of OpenStat, it'll all have been worthwhile. The Windows option is a bit of a compromise.

Thanks again folks.