I have looked everywhere for a really good opensource alternative to the big-guys like SAS, SPSS, etc. Does anyone know if there is an opensource implementation of these stats packages. I am aware of R-project, but do not know how it works. I am so used to GUI interfaces (and SAS). I have found KStat which is not yet in production and SalStat but also it has limited capabilities.

Any pointers will be helpful.
Thanks,
Smiley

Hi Smiley.
Sorry, I can not give a lot of input. I have been looking also. There are a lot of good packages out there and it takes a lot of time and money to evaluate them. I have thought about and considered packages like Matlab, Mathematica, etc. but for technical business reasons a commercial package is not attractive in the long run. There is also Octavia and others, but for various reasons I had picked SciLab from INRIA in france to explore farther. I downloaded executables to evaluate, but you can also compile the code. It appears to offer a lot, there is reasonable documentation, but there is a lot of small things not covered and it is difficult to find user groups especially in the US. The folks from Inria do respond, but it takes a while. The big reason for initially lookig at it was that there was code set up to translate Matlab scripts into Scilab scripts which would have been very worthwhile due to our collaboration with UC at Berkeley where they use Matlab. But the translations are not very good, and it takes more modifications to properly use the graphics.

I recently started looking at www.r-project.org, but I am at a loss to say if it is worthwhile pursueing this path. All of the good packages take a lot of time to evaluate and my real interest is in studying the optical physics and electronic design for manufacturing. The statistical analysis is crucial and requires work, but guidance and especially good user groups would help a lot.

Best regards,
MunichTexan

Thanks munichtexan,

I agree with you. R-project has a very strong following at some universities, and it is a solid package. It takes time to learn the syntax, but they tell me it is worth the effort, as the rewards are easy to see when you become productive. So they say. I have spent some time tryin gto figure it out, it is the closest I can get to doing complex stats (life sciences and epidemiology). Documentation is aso quite good with a strong user forum. I believe, PSPP, a pun on the SPSS package, is working on getting SPSS-like package, but is in development, and from what I see there is very little progress there. KStat is planning t incorporate r-project syntax and analytic engine, into a GUI based package, but again, it is very early to say what will become of these. I just wished I could help out, but I am not a programmer.

I will give Scilab a try and see. thanks, regards,Smiley

Hi Smiley,
Thanks for the input. We have programming capability (not myself personally) but good analysis takes at least 6 man-months on each of these programs. I have been collaborating with the CS group at Texas State in San Marcos, Texas. We are looking at starting to evaluate these packages further and appreciate any information which helps in a pre-screening of the packages.

You mentioned GUI associated with the r-project. We are just starting to look at the graphics capabilities and I wonder if you can give me some details about it. The Scilab graphics are reasonable and you are able to expand to your own graphics library if you find in the long run that the graphics are not worthwhile. What would you say about the r-project graphics capabilities? Especially the ability to slice data out of 3D data to make a 2D representation? You are able to do this with the VTK libraries, but it would be nice to be able to do it under a single package. In addition, do to the overall database complexities our general database is being setup on HDF5 from the University of Illinois at Urbana. Are you familiar with the r-project compatibility with this database?

I am also not a programmer, but I understand what I would like to see in the end, so any information is worthwhile. We primarily like staying with the tk/tcl GUI's because it allows flexibility without having to recompile and is highly compatible with different OS . When we do need to speed up graphics, we typically write the code out in C++.

MunichTexan

Do you use an emulator? If you want to try, download, compile and install the latest wine package from www.winehq.org (version 0.9.9 up to now). (After unpacking, enter the wine directory and run: ./tools/wineinstall - it should do all the things by itself

I've tested SPSS and Minitab a bit. Here's a link you can follow for read other reviews:

http://appdb.winehq.org/search.php?q=minitab

You'll find there a review about Matlab too.

SPSS 8.0 seems to install and work well. Except the diagrams, I think. Not sure about newer version, though, it seems like they are VBOX protected and do not work because of that...

About Minitab: I've tested version 12.1. As you'll read I couldn't find anything unusual, the program seemed to work just like in Windouz. I have tested a couple of functions only, I don't have the knowledge and time to test them all