What Statistics Software do you recommend?

[MakubeX]

I would like to know the softwares that you use for statistics (excluding SPSS, obviously) or statistical computations on Linux.

Thanks!

[Wes24]

I would like to know the softwares that you use for statistics (excluding SPSS, obviously) or statistical computations on Linux.

Thanks!

Don't think that there is so much available, according to my knowledge the only still actively maintained package is pspp (PSPP (http://www.gnu.org/software/pspp/pspp.html)). I could be wrong however...;)

That's one of the reasons I keep dual booting with XP: I need SPSS for University.

[Manny C]

A couple notable ones. You can download and run the open-source version of S-plus, also known as "R" on linux. You will find it packaged as R-base. Someone is currently working on getting the linux version of WinBUGS operating through R. Not promising at the moment.

I believe that a linux version of SAS is available (but non-free).

You can (I have) bought Matlab and install the statistics toolkit. Ditto for Maple.

Did you have a particular software package in mind?

[ssam]

you might find some useful info here https://wiki.ubuntu.com/UbuntuScientists

[MakubeX]

Thanks for the suggestions and recommendations!

[dudus]

Octave has some statistics. Not many, some. But it had cover my needs.

[tommie74]

I have installed SPSS 11.5 under Linux using codeweavers cxoffice 6.1.0. (commercial version of wine). I used the same botlle (virtual windwos c: disk) that I installed office 2000 on. Everything seems to work allright except for the graph functions. I can can load data, analyze it, produce tables with output etc. Only producing graphs seems problematic. The install is going smoothly, asking for a serial etc.
So if you can miss graphs, SPSS seems to work allright. The more people us SPSS with cxoffice, the bigger the chance that they will start support on it.
Greetings,
Thomas.

[akniss]

I have installed SPSS 11.5 under Linux using codeweavers cxoffice 6.1.0. (commercial version of wine). I used the same botlle (virtual windwos c: disk) that I installed office 2000 on. Everything seems to work allright except for the graph functions. I can can load data, analyze it, produce tables with output etc. Only producing graphs seems problematic. The install is going smoothly, asking for a serial etc.
So if you can miss graphs, SPSS seems to work allright. The more people us SPSS with cxoffice, the bigger the chance that they will start support on it.
Greetings,
Thomas.

Thanks for the information, but is there any particular reason you posted the exact same thing in 4 diferent threads?

http://ubuntuforums.org/showpost.php?p=3056277&postcount=7
http://ubuntuforums.org/showpost.php?p=3056249&postcount=2
http://ubuntuforums.org/showpost.php?p=3056274&postcount=3
http://ubuntuforums.org/showpost.php?p=3056276&postcount=5

And while it may be true that if more people use SPSS with crossover than it will be better supported, wouldn't it be better to support a program that actually runs natively (with functioning graphing capabilities) on linux if that is the OS of choice? Or if more people tell the makers of SPSS that they want a Linux version instead?

[zasf]

I raccomend R (package r-base). You can also find matlab for linux, if you're used to that and want to make your passage to linux easyer

[euler_fan]

I highly recommend R for anyone looking to do more "high powered" or customized work in statistics.

I've used R, SAS, and Minitab (R on both, SAS, Minitab on Windows) and while SAS can be great, I think SAS is better at set/well-defined analyses whereas R is better for custom work. R is also designed as a language/environment where the SAS macro facility left me feeling like it was an add-on. Custom scripting is much closer to the everyday in R.

[mycotropic]

I installed SAS Linux the other day and it's a monster beast creature from hell to get running as far as I can tell. I've been a SAS programmer for ~10 years and I'm actually getting a little desperate to get it working in this environment.

Oh and the instructions provided by SAS for Linux install? Every single command was incorrect using Ubuntu. All of them.

And of course it still doesn't run but I'll post an "install SAS under Ubuntu" thread as soon as I work out what the issue is.

[machoo02]

Gotta go with R. It may take a little time to learn, but there are plenty of references and tutorials out there to make this process a little easier.

[docdoc]

I really need SAS. I tried to install it a year ago but never got it to work. I then worked a bit with SAS in Virtualbox but eventually left Ubuntu again for XP.
I am a doctor and researcher and currently I haven't got the time to change to R. I have tried it a number of times but when things get complicated I am back in SAS.

If anyone has got SAS to work under Ubuntu I would very much like to hear how.

[zasf]

If anyone has got SAS to work under Ubuntu I would very much like to hear how.

I've never tried SAS on linux, but I gave a chance to Matlab on linux. I can tell you that the user interface is really ugly and buggy compared to windows.. in my opinion it is worth a try if you want to do an easy switch from windows to linux, ie first you switch os and then you switch the programs you use.

This will allow you to run your old programs within the new os, but in the long run it is much better to go on open source software.

[earlycj5]

R but I do have a computer at my disposal with RHEL4 and SAS for Linux on it.

Still I prefer R where I can use it.

[machoo02]

I really need SAS. I tried to install it a year ago but never got it to work. I then worked a bit with SAS in Virtualbox but eventually left Ubuntu again for XP.
I am a doctor and researcher and currently I haven't got the time to change to R. I have tried it a number of times but when things get complicated I am back in SAS.

If anyone has got SAS to work under Ubuntu I would very much like to hear how.

Are you trying to install the Windows version of SAS? I looked over that the Wine AppDB (http://appdb.winehq.org), and although the reports are old no one seems to have had luck installing SAS through WINE. It seems that there is a Linux version of SAS out there (see earlier posts). Is this the version you are trying?

[docdoc]

Originally Posted by machoo02:
Are you trying to install the Windows version of SAS? I looked over that the Wine AppDB, and although the reports are old no one seems to have had luck installing SAS through WINE. It seems that there is a Linux version of SAS out there (see earlier posts). Is this the version you are trying?

No, given the complexity of SAS, which may even give you a hard time installing it under XP, Wine is not an option.
I tried to install the SAS for Linux v9.13. The thing is that SAS only support Suse and RHEL. Maybe I'll give openSuse another try even though I prefer Ubuntu.

[bingouk]

R is superb for doing serious statistics and has thousands of exta packages for download from CRAN. There some useful frontends to the command line. I think JGR works in linux but the best for new users is RCMDR (http://socserv.mcmaster.ca/jfox/Misc/Rcmdr/). Not tried it for linux and use the command line anyway but it's a good way to learn the syntax.

[machoo02]

If you do decide to switch to R, the R-wiki (http://wiki.r-project.org) has a page on some sample translations betwen SAS and R code (http://wiki.r-project.org/rwiki/doku.php?id=getting-started:translations:sas2r) and several other statistics programs (http://wiki.r-project.org/rwiki/doku.php?id=getting-started:translations:translations).

[gnuman]

I just got done taking a course in SAS programming. Although SAS is very capable of handling large data sets, my heart is with R. I've removed the "limited student educational package discount rate" version and will now dedicate my free time to learn R. It's quite capable and efficient, IMO.

):P

[bingouk]

I just got done taking a course in SAS programming. Although SAS is very capable of handling large data sets, my heart is with R. I've removed the "limited student educational package discount rate" version and will now dedicate my free time to learn R. It's quite capable and efficient, IMO.

):P

I agree! SAS is the default in the medical industry like S-Plus was in academia a few years ago. For me R has overtaken both and after moving from S-Plus I was shocked by how much more is available in R. I had to do a generalised partial least squares analysis recently (with a multinomial rather than continuous response and several thousand covariates). My employers had MATLAB, SAS, S-PLUS and GENSTAT. Only R could do the job with the gpls package! And if your into bioinformatics R is the ONLY option for me!

[kstella]

I find more and more reason to love the forums every time I run a search. I'm looking for statistics software, and this thread was very helpful. Thanks, guys! :)

[plvera]

I also highly recommend R. Although the learning curve can be steep, the results are definitely worth it.

Check out the main web page and the help newsgroups.

http://www.r-project.org/

There are plenty of manuals online, but I recomend getting:

Introductory Statistics with R, Dalgaard, P. (2002)
The R Book, Crawley, M.J. (2007)

Good luck!

[zasf]

If you want to learn basic time series analysis and get started with R, I would reccomend Time Series Analysis with R - Part I (www.statoek.wiso.uni-goettingen.de/veranstaltungen/zeitreihen/sommer03/ts_r_intro.pdf -)

Matteo

[subs]

this was intriguing...

[Klipt]

Also check out the R/SPlus-Python Interface (http://www.omegahat.org/RSPython/index.html) which lets you call R functions from Python! There are some statistical packages for Python (http://www.astro.cornell.edu/staff/loredo/statpy/) but currently much more limited than R.

[dirkjot]

I have SAS successfully working in a Qemu box. This is very easy to set up and works well. Qemu is a full emulator, which allows you to boot up windows inside a window on linux. It works very well and is very fast (if you install kqemu, which is not in gutsy yet, as far as I know, but easy to install nonetheless). Tip: After setting things up, use the VNC interface over the standard SDL one.

I installed windows 2000 inside my qemu 2Gig fake harddisk and then installed SAS into it without any problems. SAS works like a charm. I have also installed SPSS so I now have all my stats programs available. With some simple setup, you can share documents between the windows and unix environment.

hope this helps,
Dirk

http://fabrice.bellard.free.fr/qemu/

[homeriq5]

There is indeed a version of SAS 9.1.3 that is available for linux, however it is not free unless it is licensed by your company/university. I was able to pick up a free copy of the linux version of SAS 9.1.3 today from my school and had it installed with only a few minor problems. The setup went relatively smoothly, but when I tried to invoke SAS in the terminal following installation it gave me an error message about not being able to write to the "Work" folder and was thus not able to open. I googled the error message and found a solution online which involved changing the location of the work folder to a location where I had read/write permissions. The interface itself is similar to that of the windows version, but for some reason the program editor, log, output, explorer, etc were all contained in separate windows instead within one SAS window. So far, I think the windows version is a lot easier to use

[earlycj5]

I've been using SAS on my openSUSE laptop lately, but not directly on it.

We have SAS 9.1 on our main Sun machine on campus that hosts our e-mail, etc.

You can easily ssh in.

%ssh -X user@host.host.whatever


Allow your X server to allow the connection from the remote host (type this on your local machine).

%xhost+


Doing this and using knetworkattach to attach to my remote account the work is pretty much seamless.

[TimMills]

I use R too, and recommend it (and Ubuntu, LaTeX, etc) to any of my fellow PhD students who will listen.

I had one mild problem installing extra packages in R - the command in R to install packages doesn't work if you are running R as a normal user (you don't have permissions).

If a package is in the repositories, no problem - Symantec is my friend.

If it's not in the repositories, the R command to install packages is

> install.packages("...")

entered during an R session.

But if you want to install a package on the system (not just in your home directory), you need super-user privileges. So rather than starting R the normal way:

$ R

you need to start it as an administrator:

$ sudo R

Then you run install.packages(), exit R, re-enter as a normal user, and all is well.

Also, I'd add Harald Baayen's book Analysing Linguistic Data: A practical introduction to statistics - almost out, but available here as a draft: http://www.ualberta.ca/~baayen/publications/BaayenCUPstats.pdf

It's written for linguists, with worked examples available as an R package, and every step explained (including exact commands) in the text.

[knattlhuber]

I really need SAS. I tried to install it a year ago but never got it to work. I then worked a bit with SAS in Virtualbox but eventually left Ubuntu again for XP.
I am a doctor and researcher and currently I haven't got the time to change to R. I have tried it a number of times but when things get complicated I am back in SAS.

If anyone has got SAS to work under Ubuntu I would very much like to hear how.

I managed to install SAS 9.1.3 way back in 2005 on an older Ubuntu version. After fixing some errors in the license file, it installed without any major problems.
However, the usability of the Linux version was so poor that I exchanged it for the Windows version within a week. E.g. the editor wouldn't do highlighting or drag&drop, carriage returns didn't work. It was all very weird. A friend of mine, who also uses SAS under Ubuntu, said that because of that he uses Vim to write/edit his syntax and then pipes the file to SAS.

[knattlhuber]

Another stats package that should be mentioned in this thread is Stata. There is a Linux version of Stata available.

While Stata is not as powerful as R or SAS, it's IMO better than SPSS and it should satisfy the needs of the average biostatistician/epidemiologist. Stata has a quite strong user community that provides a lot of extra commands/tools. It might be worth giving it a try if SAS won't install and R is too complex.

Having said that, I haven't tried Stata on Ubuntu yet. I still run my analyses on Stata for Mac OSX..

[earlycj5]

Another stats package that should be mentioned in this thread is Stata. There is a Linux version of Stata available.

While Stata is not as powerful as R or SAS, it's IMO better than SPSS and it should satisfy the needs of the average biostatistician/epidemiologist. Stata has a quite strong user community that provides a lot of extra commands/tools. It might be worth giving it a try if SAS won't install and R is too complex.

Having said that, I haven't tried Stata on Ubuntu yet. I still run my analyses on Stata for Mac OSX..

Interesting, never used Stata, when I was involved in teaching epidemiology we taught the class using R.

[gah789]

Stata works fine under several flavours of Linux that I have tried it on - including Ubuntu & OpenSuse. The structure of the program - with extensions provided by ado files, etc - means that it is only necessary for the developers to get the core compiler/interpreter right and the rest can be left to individual modules. I use it under Windows & Mac OS-X as well and haven't come across major problems switching programs from one implementation to another.

I gave up using SAS - and previously SPSS - because of its cost and inflexibility. The only reason to use either of them now would be if one had to process massive amounts of data or, in the case of SAS, for complex data manipulation. But under Linux with lots of RAM and/or multiiple processors it is possible to use the SE or MP versions of Stata for relatively large datasets.

R can probably do anything that Stata can do and more, but I find that it is a world of its own and switching from SAS is likely to be horribly slow, whereas the introductory guides to Stata are quite good and the transition is relatively painless.

[pape]

I've mostly use Stata, which I think is an excellent tool available for both Windows and Linux. The only problems with it are that it is commercial, and they annoyingly make different Stata versions incompatible with each other, keeping people in an expensive upgrade cycle (It can be 'found' from the interwebs as well...) .

Once I find the time, I will definitely learn R, but will probably remain dependent on Stata for a long time.

Also note that for doing econometrics, there is gretl, an open source package, which does most of the basic stuff (ie undergraduate econometrics).

http://gretl.sourceforge.net/

[Zimmer]

http://ubuntuforums.org/showthread.php?t=641448
see post #2
SAS doesn't automatically define a directory to use as the "WORK" library, so you have to define it yourself when you start the program.