SOFA Statistics has now become the most downloaded general purpose statistics application on SourceForge, having passed the 40,000 download mark. There are plenty of new features forthcoming, and every effort will be made to add them without compromising the commitment to ease of use, learn as you go, and beautiful output.
Archive for the ‘general’ Category
For line charts, users can now add trend lines and smoothed lines (using a weighted rolling average).
And histograms can now be produced with a normal distribution curve overlaid.
Other improvements include:
- Histograms display better when lots of bins – wider chart and smaller font for x-axis.
- Simplified layout of project selection form and changed preferences form.
- Clearer labels when users are selecting a range of groups rather than discrete groups e.g. in ANOVA.
- Added medians etc to the Wilcoxon Signed Ranks output.//
- An early error message directs the user to the help page on “SOFA Won’t Start”. It also includes connection details in the error message if unable to connect to the default database.
- Only delays next “What’s up” connection date if it successfully gets through.
There have also been some useful bug fixes:
- Bug fix when importing from ODS any data which could potentially be a year e.g. 2631.
- Report quality histograms no longer leave out values falling right on the upper limit.
- Two-tailed vs one-tailed corrections in output display.
- Fixed problem with table names (esp safetblname) surviving test for SQLite validity and breaking future tests.
I suspect that downloading SOFA Statistics is, for most users, the last contact they have with the project or the community. The latest version (1.0.2) addresses this. Approximately 2 weeks after the first use, and every couple of months thereafter, SOFA Statistics pops up a “What’s Happening” web page within the application. This page has links to help, quotes, latest SOFA Statistics news, resources, featured statistics videos etc. I hope it will be of value to users, and help them get the most out of SOFA Statistics and the training etc resources available. This aspect of SOFA is intended to be relatively unobtrusive and it can be closed immediately if so desired with its close button.
Other changes include:
- Added explanation of Mann-Whitney U to output, added degrees of freedom to output wherever possible, added footnotes for p values where lacking, and ensured all p values are to 3 decimal points unless under 0.001 in which case ‘< 0.001' is displayed.
- Better help messages if early problem with comtypes (Windows only) or library dates.
Bug fixes include:
- Fixed bug when reporting on datetime fields as rows or columns in report tables in MySQL, PostgreSQL, Access, and SQL Server.
- Fixed minor bug which meant MySQL importing failed on Macs because it didn’t recognise the platform.
Work will begin soon on the versions leading up to 1.1.
SOFA Statistics is 100% free and open source but affordable commercial support is available to those who would like that here. The latest version is the culmination of several years work and we hope you like it!
Installation note if upgrading: if upgrading from 0.9.24 or below, you will need to rename your /home/username/sofa folder to /home/username/sofastats if you wish to work with old data or settings. If you are using Ubuntu, you must also manually uninstall previous versions of SOFA first so that “sofastats” can overwrite the application icon put there originally by “sofa”.
One of the biggest changes this release is behind the scenes – the folders SOFA Statistics is stored in have changed from “sofa” to “sofastats” to make them more distinct (there are other sofa programs ;-)). This was something important to get done before the version 1.0 release. Although this is a good change for the long-term it means users upgrading from 0.9.24 and below will need to rename their /home/username/sofa folder to /home/username/sofastats if they wish to work with old data or settings. For Ubuntu users, it is also important to manually uninstall previous versions of SOFA first so that “sofastats” can overwrite the application icon put there originally by “sofa”.
More noticeable, is the change to version upgrade checking. Auto-checking for any version upgrades is now the default. The goal is to help users stay up-to-date with bug fixes, improvements etc. Users can disable this feature by selecting the “No checking” option, or only auto-check for major upgrades. This feature only reads the contents of a small text file on the sofastatistics.com domain containing the latest version number.
Other changes include:
- SOFA suggests names more likely to be valid when user selects a file for importing.
- Can now import csv files with a txt extension.
- MySQL and PostgreSQL can now accept blank passwords.
- Better message for user if comtypes problem.
- Windows and Mac versions will no longer accept being launched with Python 2.7 and will give useful messages about fixing the problem. For Windows, the importance of installing the site-packages to python26 rather than python27 is explained.
- Mac version uses new, easier-to-install library to connect to MySQL databases.
The main series of bug fixes relate to recoding data and making new data tables or redesigning them.
- Fixed numerous recode bugs – mostly to do with application of recoded value labels.
- Fixed bug when making a new table with an existing name. Now stays open to give user choice as to what to do next e.g. change table name.
- Fixed misc bugs when deleting/adding/renaming tables while a database other than the default SOFA database is selected.
- Fixed bug running frequency table using MySQL where the category items are long integers.
- Fixed bug where Data List report selected and then the database is changed.
Ideally, the next release should be version 1.0.0
SOFA Statistics has now been downloaded over 30,000 times and the pace has accelerated in recent months. This is encouraging for the project as preparations are made for the version 1.0 release.
Version 1.0 of SOFA Statistics is about to be released. This is a big opportunity to introduce SOFA Statistics to a much wider audience.
SOFA Statistics is already receiving positive attention in the open source and Linux community. Linux Journal, for example, recently reviewed SOFA Statistics and ended the piece with “My hope is that this program becomes an adopted industry standard of sorts, mentioned in everyday conversation by organization workers the world over …” (Linux Journal Jan 2011, Issue 201, pp.40-41). This is fantastic support for the project, and SOFA was also fortunate enough to get on the cover of the Ubuntu magazine, Full Circle Magazine:
The challenge is to get into the wider IT press and beyond so that people looking for a stats/reporting package automatically include SOFA Statistics as one of their options.
Do you have any suggestions? We welcome comments on this topic. A version 1.0 release is an opportunity that shouldn’t be missed. You can either post a message or email me at email@example.com
A key goal of SOFA Statistics is to help users “learn as they go”. The latest version takes a further step in that direction by integrating on-line help into all the statistical tests. In the example below, clicking on the “Help” button in the Chi Square dialog opens the appropriate page in the client’s web browser.
Version 0.9.24 also makes it easier for users to select the right test, and feel confident they have made an appropriate choice, by adding tips for any tests previously lacking any.
The other change is that passwords are now hidden when configuring database connections in projects.
There has been an important bug fix for a problem introduced in 0.9.23 which prevented Excel importing.
Other bug fixes have included:
- PostgreSQL and MS SQL Server now cope with databases which are listed but where a connection is not possible e.g. because of permissions.
- Fixed minor bug in sqlite table name test – now commits when making table and before dropping it so a chance to succeed or fail on individual step. Otherwise, net effect is no change (table added and removed) so no chance to fail.
- Safer reading and executing project and variable detail code in Windows (change new lines to required sort).
- Now possible to run test code in Windows (change new lines to required sort).
- Fixed bug when refreshing default sofa tables but not in the sofa sqlite connection.
This release brings SOFA another step closer to the version 1.0 release expected early in 2011.
SOFA Statistics version 0.9.23 makes importing data from spreadsheets (and csv) much more robust. In particular, it copes better with inconsistent data types in a single field and give the user a choice as to which to use:
The latest version also has a much more reliable method of identifying and extracting date data from ods (OpenOffice Calc, Gnumeric, Google Docs spreadsheets).
Other changes include:
- If no change to table design when clicking Update, get a message that there was no change instead of no response.
- When opening data tables, the read only checkbox is ticked or unticked according to whether showing the demonstration table or not when in the default sofa database.
- Now handles PostgreSQL databases with different schemas.
- When run in console, SOFA produces much more useful input on progress during initial stages and on any error encountered.
- Height setting algorithm for data editing/display grid now copes with taller e.g. double height, taskbars in Windows.
- More reliable method used for identifying the folder location of SOFA.
There are also numerous bug fixes:
- All lines in multi-line text cells are now imported from ods spreasdsheets correctly.
- Errors when getting project settings file at start return useful message and don’t fail because tried to use msgbox before the GUI application is running.
- Better handling of errors when identifying date format. Falls back to a reasonable default.
- Fixed error preventing message being shown when unable to change database details e.g. selecting a different db in the drop-down.
- Fixed bug in legend of auxiliary clustered bar charts when long value labels. Now has automatic line splitting.
- Correct initial display of csv data when importing if missing cells in some rows.
- Majorly reduced impact of bug when creating report tables using numbers with more decimal points than python displays when converting to text e.g. into an SQL statement ready for execution.
- CSV importing now provides useful error messages if actually an Excel spreadsheet saved as a csv with Excel features preserved (i.e. a faulty csv file). Also copes better with faulty delimiters being manually supplied.