Stats and Graphing for the Mathematically disinclined

[gunksta]

I installed OpenStat. I created a simple, fake dataset and played around with it. It's UI seems pretty friendly and easy to use.

[hzambran_cl]

It is apparent that although powerful and free applications such as R and GnuPlot may be a godsend to some it seems highly unlikely that they will be adopted by the masses who may be more comfortable using SPSS and Sigmaplot or similar programs.

I disagree with this opinion, and for supporting my disagreement, I would suggest to read the following articles on the New York Times:

http://www.nytimes.com/2009/01/07/te...07program.html

http://bits.blogs.nytimes.com/2009/0...u-ready-for-r/

Obviously, at the end anyone is free to choose the "best" tool for a particular job (citing part of the nice 'gunksta' message), and free to define what "best" means for himself...

[ahmatti]

For easy plotting of data and simple curve fitting see kst http://kst.kde.org/

[BramWillemsen]

I've always used xmgrace for plotting. It's quite a wonderful tool and not too hard to learn. I've always used similar types of input so i maybe not be too familiar with all the options of the program, but for plotting its absolutely amazing. i always used two column output files (x,y). You take these as input and xmgrace plots it for you. You can use as many input files as you want in the same graph.

You can adjust colours, symbols for each point and many other options. The output looks pretty neat too. I'm currently working on my bachelor scription and these two figures show some of my model output. Turning model output into figures is really easy by killing the previous data and inserting data from new 2-column files. All the settings like labels and axes remain the same then.

http://i35.photobucket.com/albums/d1...igure3temp.png
http://i35.photobucket.com/albums/d1...igure6temp.png

Hope this helps you a bit.

[sdbrogan]

I hadn't heard of OpenStat, so decided to have a dekko. I tried multiple regression on the Longley test data (a tricky case that can cause weak algorithms to fail) & checked the results against those certified by NIST; in short it gave the wrong answers. Did you notice when you start it up you're warned "You should attempt to verify the accuracy of results with hand-calculated examples or other data of known solution" ? It's a nice piece of software, & could be handy for students, but it's surely not the substitute for SPSS for researchers you asked for.
I think your options are (1) buying SPSS (it's available for Linux OSs), (2) waiting for PSPP to have all the functionality you need, or (3) using R (which passes the Longley test of course).
I don't think the difficulty of using R should be exaggerated. Even using it from the command line any common test can be carried out in one command - it's a high-level language. Cut & paste from a crib sheet if you don't use it often enough to remember the syntax. R Commander is another GUI front end you could try if you don't like RKWard - it's not particularly swish, but as easy to get the hang of as any other GUI.
The Longley data is at
http://www.itl.nist.gov/div898/strd/.../Longley.shtml

[macotto]

A speadsheet program (if it can graph what you want . . . might be a big if) is probably the `easiest' or at least the most gui'ed plot maker.

[sdbrogan]

By the way, PSPP also gives the right answers for this regression.

[lethalfang]

I was wondering what the best programs to use for those social, behavioral, and heath researchers who have a weak math and computer science background.

I am a PhD student in audiology and am using OpenOffice as my word processor with Bibus as my referencer. This combo seems to be workin good so far.

For stats, I have installed OpenStat through Wine which seems like a great alternative to SPSS. PSPP has a nice interface but is far too limited. The R GUI, RKWard, looks nice but it just doesnt seem quite as intuitive for persons with an SPSS background.

For graphing, I have been using OpenOffice. I may try Gnumeric but I'm guessing its graphs will look about the same. I liked RLPlot but it is kinda buggy. I havent had time to fully explore Grace yet.

I realize that programs like R and GnuPlot are superior for those that are proficient with them but I am more interested in hearing about experiences with programs that have nice, intuitive GUIs.

Any thoughts or comments?

You can try GNU Octave along with its GUI frontend QtOctave. The two are designed to be compatible with MATLAB, which is a very powerful numeric program. It makes nice graphs and stuff.... don't know if it fits your needs.

[nanonils]

Unfortunately, I will have to use JMP8 to share files with collaborators. Because I'm still fairly new to Linux I have a question regarding the installation location: should that be /home or /opt as suggested by JMP8?

I am the administrator of my computer and occasionally would like to do a clean Ubuntu install, keeping /home on a separate partition.

Is it OK to follow the default location and then just copy this back or would you simply install it on /home? Will links break after a clean install?

[gunksta]

Unfortunately, I will have to use JMP8 to share files with collaborators. Because I'm still fairly new to Linux I have a question regarding the installation location: should that be /home or /opt as suggested by JMP8?

I am the administrator of my computer and occasionally would like to do a clean Ubuntu install, keeping /home on a separate partition.

Is it OK to follow the default location and then just copy this back or would you simply install it on /home? Will links break after a clean install?

nanonils: If this doesn't answer your question, consider starting a new thread. You'll probably get more help that way. I've never used JMP 8, but your questions are pretty generic.

I would install JMP 8 to /opt, as per the directions. I don't like to have executable in my home directory. If you do choose to install to /home, I would recommend doing the installation that way up front. Copying _could_ break links, depending on how the installer works.

The reason I recommend /opt is because clean installs are rarely necessary with Ubuntu. (although they can be fun.) If you do a clean install, you might as well reinstall JMP 8 too. It shouldn't take that long. The big risk with stuff like JMP 8, is that in the future you could have problems with things like libc. Open source programs, like R, PSPP, etc. are recompiled for each Ubuntu release with the latest and greatest system libs, graphical toolkits, etc. JMP 8 is not, since it is proprietary. It is, in effect, stuck in time like a fossil. If something that JMP 8 depends on changes dramatically (breaks API/ABI compatibility) it may not work in the future. I recommend that you have a non-critical test-machine so you can test upgrades or you could upgrade and "break" JMP 8 without meaning to.