Statistical programs: SPSS16 and STATA 10

Sigg3.net · **Posted:** Fri Jun 06, 2008 12:59 pm

My boss's best argument not to switch to linux has been those two applications which we really need. The open source PSPP is unstable, and although I've read that Ubuntu users have managed to run SPSS16 I must know positively that it works. I'm planning to do some wine tests to see if that could do the trick. Just fishing for other people's experiences.

Both STATA10 and SPSS16 have linux versions.
http://www.stata.com/products/opsys.html
http://www.spss.com/spss/system_req.htm (Only supported on Red Hat® Enterprise Linux 4 Desktop and Debian® 3.1 atm.)
but at the moment we only have versions for Windows (and they are really expensive).

Have anyone tried these apps in Zenwalk? I would love to switch to ZW before someone upstairs says: "We should all move to Vista now."

Btw, just noticed The R Project. Anyone here familiar with it?

markc · **Posted:** Fri Jun 06, 2008 4:57 pm

R is in /extra and is maintained by gapan

gapan · **Posted:** Fri Jun 06, 2008 5:38 pm

I'm quite surprised stata and SPSS have linux versions, I've never seen those before. If they are providing linux versions, I guess they do work.

SPSS on wine simply doesn't work. Stata, I haven't tried at all. But I'm happily using Minitab 15 under wine with very few problems. I find the SPSS interface ugly and completely unintuitive, I only end up using it if nothing else is around.

Do you mind sharing what you do with stata and spss in your company exactly? I'm asking because I know lots of people that believe that a mean comparison using a t-test or ANOVA or a regression calculation for example can only be done with spss and nothing else.

R is Linus Torvald. You can do anything with it. But it has a very steep learning curve. You will have to go through the manual very carefully. And then you will have to go through the manual again. And again. But in the end it pays out, especially if you have huge datasets.

digger95 · **Joined:** Mon Apr 28, 2008 10:15 pm **Posts:** 21

I did the 14-day free trial of SPSS 16 for Linux. It installed just fine on Zenwalk, however it is java-based and it is sloooooow.

gapan · **Posted:** Fri Jun 06, 2008 6:02 pm

I was talking about SPSS 15 and previous versions, haven't tried 16. Java... so that's how they support multiple platforms then. I haven't tried but it has to be even slower than it was before.

digger95 · **Joined:** Mon Apr 28, 2008 10:15 pm **Posts:** 21

Yeah, SPSS 16 is their first Linux version. I googled the problem quite a bit after installing it, and apparently the new version is just painfully slow across all platforms, not just Linux. I understand their reasoning for using java but the software is dang near unusable now. SPSS 13 was pretty fast on this same machine under Windows.

Sigg3.net · **Posted:** Mon Jun 09, 2008 10:17 am

I'm a philosopher, not a statistician, so you'll have to find out from our website: http://www.fafo.no/ais/index.htm
What I do know is that we use a lot of big datasets and complex syntaxes... and that our boss with version 16 also wants me to install the Python addons.

Digger95, some of it is due to graphical errors and can be solved, but *NIX and Win users alike are complaining over the reduced speed related to new output files, so I guess they'll have to patch it pretty soon.
I have all the versions since 11, so we can always roll-back (though there are some version dependent incompatibilities).

Also I read this at the ubuntu forums:

Quote:

The only problem is that SPSS for linux is incomplete, the non-parametric tests only work in windows.

gapan · **Posted:** Mon Jun 09, 2008 6:14 pm

I can guarantee you that you can do everything you want with R, but it will require a lot of patience to get to know how it works. Once you know how to work with your data with R, it will be a lot faster to work with it and get results than with spss. Sort of like comparing clicking around to make complex tasks in a PC and learning to create a script to do it all for you. Depends on how much time you can spend learning a completely new thing I guess.

Sigg3.net · **Posted:** Wed Jun 11, 2008 10:05 am

It's more of a programming language for statistics than a big engine for running syntaxes?

gapan · **Posted:** Wed Jun 11, 2008 11:01 am

Well, yes, it needs a lot of getting used to, but it pays off.

daacosta · **Joined:** Sat Aug 30, 2008 8:19 pm **Posts:** 4

Sigg3.net wrote:

Btw, just noticed The R Project. Anyone here familiar with it?

Yes, I am familiar and I believe that this is the de facto standard in the scientific community. R has everything you need and from what I have seen it is in the repositories (R-2.7.0) and can be installed easily (already did) R can open SAS files as well as Minitab's from what I have seen so it is really more than you'd ever expect. Yes, it is a "programming language" but it's got all you need for your statistics needs.

I have just one small question: Since I have already installed R-2.7.0, can I simply install R-2.7.2? I'd like to have the latest version of R but if I can't I will be OK

gapan · **Posted:** Sat Aug 30, 2008 9:22 pm

daacosta wrote:

I have just one small question: Since I have already installed R-2.7.0, can I simply install R-2.7.2? I'd like to have the latest version of R but if I can't I will be OK

New R package on the way to the snapshot repos: http://pnboy.pinguix.com/gapan/R/R-2.7.2-i486-54.1.tgz

daacosta · **Joined:** Sat Aug 30, 2008 8:19 pm **Posts:** 4

gapan wrote:

daacosta wrote:

I have just one small question: Since I have already installed R-2.7.0, can I simply install R-2.7.2? I'd like to have the latest version of R but if I can't I will be OK

New R package on the way to the snapshot repos: http://pnboy.pinguix.com/gapan/R/R-2.7.2-i486-54.1.tgz

Awesome! So I will not have to install it from source... I don't mind doing it though...

rancidcow · **Joined:** Mon Sep 01, 2008 6:01 am **Posts:** 1

I'm suprised to hear that you've had problems with PSPP. I can only assume that you were using a very early version. Recent versions of PSPP are particularly stable, have an intuitive GUI which will be very familiar to SPSS users and is a great free replacement of SPSS when it comes to reading syntax and data files.

True, it can't yet do all the statistical procedures of SPSS, but the vast majority of those from the base version are there. Also, in my experience, PSPP is much faster than SPSS when it comes to processing large data files.

gapan · **Posted:** Mon Sep 01, 2008 9:48 pm

I didn't know that pspp had a graphical interface. Thanks for the info rancidcow!

Code:

** (psppire:464): CRITICAL **: g_sheet_column_start_pixel: assertion `col < g_sheet_column_get_column_count(geo)' failed

Anyway, pspp is rather basic, it only has some descriptive statistics options, t-tests are only 2-tailed, ANOVA is only one-way, regression is only linear and that's about it. I can't seem to find how to make any graphs with it. I can't even copy more than one cell at a time to paste it somewhere else. It looks very promising though.

Statistical programs: SPSS16 and STATA 10

Who is online