Quad Shot of Ubuntu
Very interesting and impressive! That's my reaction to your video demonstration, which I just watched.Originally Posted by grantbuntu
I taught basic statistics at the university level for 31 years till retiring 6 years ago. I'd have loved to have SOFA Statistics for my students (and myself) to use; and I'll be recommending it to some colleagues who are still actively teaching and researching.
Last edited by rewyllys; July 29th, 2009 at 05:53 PM.
5 Cups of Ubuntu
@gunksta - thanks very much for the positive feedback.
Re: postgresql, I might have a look at that in a week or so. The Python support seems good and I have a pretty standard approach going now for each database engine plugin.
Re: statistical validation, I have started building a unit testing framework but that is still at its very early stages. Clearly, that will need to be extensive (and visible to the community) before the 1.0 release. I will definitely look at using NIST datasets as part of that (http://www.itl.nist.gov/div898/strd/general/main.html).
Re: large datasets, I expect the only limit will be Python, the database in question, and user hardware. With the report tables, for example, the data is gathered by a series of SQL statements. The program is then only tasked with displaying it. I put a limit on the number of cells you could have in a useful table. I set it quite high so no-one should bump into it unless they make a mistake and select an inappropriate variable e.g. phone number.
Re: Ubuntu repos, that is a great idea. But I would want to wait until everything is stable enough. Development is still rapid and I love the freedom to fix mistakes and improve (break) APIs. The project code is hosted at Launchpad so I expect to pursue the Ubuntu repo path later on.
Re: business plan. The plan is to provide a permanently open source core (see Open-Core Licensing (http://blogs.the451group.com/opensou...ore-licensing/). This core may even expand over time as my proprietary addons (none at present) move into the core and more advanced add-ons are created. The independence of the underlying Python code from the GUI is the key to allowing report automation. The fact that everything is displayed in HTML, even internally, implies corporate intranet and internet possibilities, which is probably where an income stream can be found.
Having said all that, I am motivated by two other things.
1) To make a product that I want to use myself in my work as an analyst. Unlike proprietary products I have worked with before, I love the ability to fix any problems I find and make my working experience more pleasant. I smoothed off several rough edges in the latest 0.8.4 release for example (https://sourceforge.net/projects/sof...l.deb/download) - see if you notice any differences. You can cheat by reading http://www.sofastatistics.com/blog/?p=97
2) To provide a gift to the world (Open Source 101). I would be delighted if school students, business analysts etc in places as far flung as Ghana, Scotland, Bolivia, Germany etc were using my program and benefiting from it.
If you find any bugs, please report them on the SOFA Statistics Launchpad site. The more the merrier. Details here - http://www.sofastatistics.com/blog/?p=92
5 Cups of Ubuntu
@rewyllys - I would really appreciate input from people with your sort of background as I further develop SOFA Statistics. I will set up a project-specific bulletin board once the project gains a little more momentum. The area I am most keen on extending will seek to address two issues:
1) which test should I use? Is my data normal enough? What does normal even mean? Can the program show me at the point of my decision making based on my actual data with mini-charts, quick subsidiary analyses, and other helpful visualisations? This is an area I will want lots of feedback on. The technology is readily available to do wonderful things in this area.
2) what do the results mean? Statistical significance is one factor, but what about the oomph of the effect? Is the difference being measured clinically/ economically/ etc significant. A statistics program can't tell me the answer but it can encourage me to ask the question and provide supporting information beyond p values etc. What are some of the next steps I should take e.g. Pearson's Chi Square was significant but which interactions were important?
Finally, if you like the program, please feel free to review it at https://sourceforge.net/projects/sofastatistics/. There are no reviews yet and that would be really good. NB to test the best available release (0.8.4) which I released last night http://www.sofastatistics.com/downloads.php.
Last edited by grantbuntu; August 1st, 2009 at 12:37 AM. Reason: Use standard approach to indicating who response is to
5 Cups of Ubuntu
@ugm6hr - I am hoping to use Raphael (http://raphaeljs.com/) for the output charting but a lot depends on what is available and when. See http://groups.google.com/group/rapha...1341b0ad26de16
Re: netbooks, I understand the issue but it may take me a while to resolve it. I am sometimes trying to cram a lot into one screen.
First Cup of Ubuntu
If you want a good answer to this question, you need to be specific about your needs. Be sure to tell which of the following factors are important to you:
- Ease of learning
- Quality of help files
- Extensibility/programmability
- Data entry and validation
- Data manipulation
Way Too Much Ubuntu
Just curious: Any specific reason for not using matplotlib http://matplotlib.sourceforge.net/?
5 Cups of Ubuntu
At this stage, I intend to use Matplotlib for the auxiliary graphing and RaphaelJS for the final output graphing.
By auxiliary graphing, I mean all the smaller charts and visualisations that help a user select the appropriate test. E.g. the charts showing if a distribution if adequately normal.
By output graphing I mean presentation-ready output e.g. the findings. The emphasis for final output is on bling or eye candy, albeit not at the expense of accuracy.
In Candy, Community, Comfort, Credibility etc I explain some of the philosophy behind SOFA Statistics and the niche I am aiming for.
Just Give Me the Beans!
Sofa does look very nice but may not be sufficient for many users desiring more complex analyses. On the other hand, R, may present too large of a learning curve for some. In these instances, a nice SPSS alternative is OpenStat.
http://statpages.org/miller/openstat/
It's pretty much updated monthly and has a huge array of statistical procedures with ample documentation available. I believe it is maintained by one man but it is very nice. However, it does need to be installed through Wine to use current versions. But it does run perfectly through Wine. The graphs are not very pretty but other software would be better suited for that function anyways.
Just Give Me the Beans!
Last time I checked, OpenStat's regression routine gave wrong answers on some reference data from NIST.
http://ubuntuforums.org/showpost.php...9&postcount=15
I think R is pretty much the only fully-fledged, general-purpose, free statistical software available at the moment.
Last edited by sdbrogan; August 14th, 2009 at 10:50 AM. Reason: afterthought
Grande Half-n-Half Cinnamon Ubuntu
I would really like to see SOFA and PSPP both grow into mature statistical environments for Linux.
R, SOFA and PSPP are each unique projects and will make Linux an even more appropriate environment for research and data analysis.
I would love to try out SOFA right now, but it looks like I would have to upgrade wxwidgets. I run a couple of programs that depend on this toolkit. I tend to be a little gun-shy about upgrading base level libraries and toolkits. I'll compile and install bleeding edge programs. If they break, so be it. But, I don't like to mess with the lower-level stuff unless it is absolutely necessary.
Right now I'm waiting for Karmic to get stable enough to justify an upgrade. Once I do that, I think I'll be able to compile SOFA with the default wxwidgets (correct me if I am wrong on this). I would love to help get SOFA into the Ubuntu repos in time for Karmic+1. With the current rate of improvement, SOFA should an advanced tool by then. If I can find the time to help, I might even get to learn more Python!
Please Insert Funny Statement Here.
Posting Permissions
Adv Reply
Bookmarks