I should have added, by the way, that no program should be blamed because it doesn't work in Vista. Most of mine don't, and that was one of my reasons for switching to Linux as my first-choice option. It does virtually everything I need and I automatically boot from my Linux partition these days.

Thanks for the interesting review. Disappointing there appears to be no top-quality GUI-based linux app in this area.

Out of curiosity I downloaded Salstat (source code package salstat.20031022.tar.gz) and it ran first time, putting up a neat spreadsheet-type interface. I have Quantian installed on this PC, which comes with a good range of packages so all dependencies must have been covered.

In fact, looking at that link I see a lot of R stuff has been added to Quantian in the last release including 'the award-winnning JGR Java GUI' and the 'Rpad interactive web interface'. I have no idea what these do. They are not in the release I have.

I'm delighted to report that openstat - now known as OS4 - does run well under Wine.

Have you looked at Maple? It isn't cheap, but it runs in Linux and it'll do pretty much anything.

Maple? I haven't heard of it until now, but I'll check it out. When you say 'isn't cheap' I hope you're not talking in terms of SPSS or SAS pricing! I'm sure you're not. And if it does produce a decent (label-editable) graphical output I'm interested. Have you a web address or can I just google it?

For others who've been following this thread, I think I should provide a further update.

openstat is a real beaut. I couldn't find a Linux download link that worked but the OS4 Windows version will run under wine. The guy behind it is called Dr Bill Millar (strangely enough, he has the same name as my statistics lecturer at university in Northern Ireland) and it's a hobby for him.

The guy is brilliant, because not only is the program terrific but he's also written a huge pdf (totally open-source) that would put some 'standard' statistics texts to shame.

If you're interested, ask and I'll give you the links.

He's recently upgraded (yes, in 2007) to Stats4U, a reincarnation of openstat and it's a cracker (that's Northern Ireland speak for very good). It's unbelievable that somebody is doing this as a hobby.

The only drawback is that it's a Windows program through and through. It runs beautifully in Vista. It installs in wine on Linux but doesn't run, unlike OS4.

Now, if there's a keen coder out there, is there any chance you might look at his source and do a .deb package for those of us who don't do too much wine?

If not, have you any ideas on how to get the program running under wine without difficulty.

As the man said, statisticians make 'mean' lovers...

Whether anyone can get Stats4U running on linux without major effort probably depends on whether Borland supply linux-compatible development tools similar to the windows ones used to build this version. The GUI part of the source code makes little sense outside the Borland C++ development environment. And without the GUI you might as well be using R

I don't think Maple will help you much. It is mainly a command-line/scripting type package like R but aimed at algebra and calculus rather than statistics.

Except for the standard deviations.

Maple does symbolic math. It solves differential equations symbolically, does integration, can do a LOT of different kinds of plots.

It is quite a formidable package and will easily handle any statistics you throw at it.

http://www.maplesoft.com/

By way of (another) update, I did finally get Stats4U running under wine, but it's not something I would recommend. Way too flaky. I couldn't find the option for labelling variable values - a little bit of a disadvantage.

Even worse, it's inclined to hang at every opportunity, trashed my bottom menu bar and froze Gnome. I had to use recovery mode to go in and delete every trace of it.

Soooo, openstat (OS4) is a good option under wine, but leave Stats4U alone. Bill Miller doesn't seem to have written any Linux code for a while, but I'll maybe contact him on his link and see if he's any plans to port Stats4U across.

Whoops!!! I just looked at my original post and see that something got hosed up. I'm sorry about that. Those " should be " so the correct syntax would beEntering data with Rcmdr is similar to entering data in a spreadsheet.

Speaking as a statistician, I wouldn't recommend using Maple. Not that there's anything wrong with it, but it isn't a stats package and won't handle many, even simple, statistics problems "out of the box." Maple and others like MATLAB are really geared towards a different branch of mathematics. On the other hand, R, SAS, SSPS, etc. are geared towards statistics and won't handle many calculus problems "out of the box." Just my two cents based on my experience with these various packages.

weibullguy, I followed your amended code and Rcmdr is running beautifully. I've imported datasets directly from SPSS with no problems and it's looking good.

I'm really grateful for your help on this and I hope you don't mind but I've got a question or two. One is how to save graphic output into another package, such as OpenOffice or even the Gimp.

It's probably really simple, but I just need a pointer in the right direction.

I'll not be on linuxquestions for a day or two because I graduate with my PhD tomorrow and there's plenty of stuff that doesn't involve statistics to be dealt with at the moment.

Thanks again to everyone for their help on this thread - especially weibullguy - and I hope it's helped others who need to do statistical analysis on Linux platforms.

Right, I'm away to graduate but I'll be back.

Here is a quick example that reads a dataset, does some exciting (but simple) analysis, saves the output to a file and the plot as a png image.

Quick example --> http://webpages.charter.net/weibullg...nrua_example.R
Example data --> http://webpages.charter.net/weibullg...xampledata.dat
Output plot --> http://webpages.charter.net/weibullg...xampleplot.png
You need this for the quick example --> http://webpages.charter.net/weibullg...guyfunctions.R

This is the section of the script that saves the plot as an image.Download the full reference manual and look at chapter 3 (grDeveices). R will produce PDF, postscript, png, jpeg, pictex (if you use LaTeX), xfig, X11, quartz, bitmap, and probably others directly. Of course, you can save the non-graphical output as unformated or formatted output as well.

Whilst we have a 'statistics' thread going I must pass on an excellent tip I found (and not by accident). I hope it will encourage more people to use Calc. The title is;
'Statistical Analysis with Excel for Dummies' by Joseph Schmuller

I use it as a guide for Calc which I use a lot. Unfortunately, there is no ANOVA in Calc (hopefully YET) but Calc and Excel have a lot of common ground. The book is recommended because the author has a remarkable gift to get straight to the substance of a problem, in what is, quite often, rough going in many books.

Hi folks,

I've finally recovered from graduation (a week ago today!) and am starting to get my head round some questionnaires that I have to code as a matter of urgency.

Really, this is probably a thank-you post to wind up the thread. I'm grateful to everyone for providing so much advice on statistics in Linux.

I've learned a lot about how helpful the Linux community is in this thread. And I've learned that I can do my analyses in Linux.

As someone who uses statistics as part of a broader research picture, I'd just like to sum up for those who might google the thread at a later date.

As I see it, there are two main options for statistical analysis in Linux. R and openstat. R is heavy-going for someone used to GUI but the Rcmdr library makes it an absolute pleasure to use. The fact that it can import datasets directly from SPSS etc means that it is simple yet powerful.

That said, I can understand why weibullguy suggested that I'd find the value of syntax important as well. Most part-time statisticians seem to forget that even SPSS is really syntax hiding under a GUI.

My advice is to look not just at the graphical/tabular output but to check out the syntax that produced it in R in the first place. There's a lot to learn from that window. And a lot worth saving for future use.

At a more basic level, I'd recommend openstat, a lovely little GUI program which will take care of most basic analyses in an environment that big-name package users are familiar with.

It's not perfect, but it's good. The creator is continually upgrading and his latest variant is Stats4U. It needs a bit of refinement yet, for it's crashy in Windows and in Linux under Wine at the moment.

It's worth watching, though, for the guy is really on-the-ball in terms of what part-time statisticians need.

There are other packages out there too. Many aren't written for Linux, but will run under wine. Like Winidams. Salstat is another option.

As someone who's serious about statistics, though, I'd have to recommend R. The Rcmdr interface and the SPSS import (I'm sure SAS, Minitab and the rest are supported too) mean ultimate compatibility without having to worry too much about CSVs, tab delimiters and the like.

By way of conclusion (and repetition, perhaps), I'd like to thank everyone who offered so much help in this thread.

It goes to prove that in Linux we really do care.

Enjoyed the thread. Thanks to all and especially weibullguy.

Just a final final word (from thread-starter) on this thread.

I've found that R (with Rcmdr) will do everything I need in terms of statistical analysis.

It's not totally straightforward for someone used to SPSS but the answer is to keep an eye on the syntax window when you do something in the GUI.

Edit the script, change the variable name and you've got whatever you want (yes, statistical nonsense if that's what you're looking for).

My advice is simple: do whatever test you want with whatever variables are available, even if they're rubbish variables.

Check the syntax to find where to put your 'real' variables
, edit/substitute and submit your amended code. It's as easy as that.

I really must thank all who took part in this thread, for the initial google results on "Linux and statistics" looked pretty miserable.

Salstat looked good, Openstat delivered basic functionality but R does all that I need and I know that it can go a lot further than anova and chi-square.

Major-league thanks to weibullguy who will, I hope, respond to my occasional "what next?" post in General.

R isn't complicated at all. It's just a matter of sorting the data u need from an imported XCel, SPSS or Gnumeric. Ok, creating a new dataset in R isn't going 2be a lot of fun but that's true of any package.

I've entrusted some of my critical data to R.

If I'm right, that compressed data will come safely out the other endfd