I was lucky enough to get interviewed by FLOSS for Science. Check it out
SOFA has been nominated for The People’s Choice Award as part of the New Zealand Open Source Awards. I would really love it if as many people as possible voted for SOFA at The People’s Choice Award. Tell your friends; spread the word.
- The latest SOFA, version 1.2.2, lets you import from Excel XLSX files (previously, Excel files had to be in the XLS format).
- Normality analyses can be included in reports, saved as output etc.
- And there is support for CUBRID databases. CUBRID is an open source relational database highly optimized for Web Applications. That brings to 6 the total number of SQL-type databases that SOFA can directly link to.
A few bugs were also fixed:
- Restored standard deviation option to row stats report tables.
- Fixed bug which meant row % was appended multiple times to config col dialog (until session closed).
- Restored PostgreSQL functionality by fixing faulty psycopg import statement.
The latest version of SOFA Statistics makes it easy (via a plug-in) to export to spreadsheet.
Additionally, there have been some important bug fixes – mainly for bugs which snuck in during the major change to 1.2.0.
- Fixed nasty bug breaking demo report tables. A casualty of the changes to the independent display of titles from report tables in 1.2.0 so wide titles didn’t mean wide table cells.
- Fixed bug with display of percent symbols in report tables. A missing not was the culprit – another casualty of the big refactoring for 1.2.0.
- Fixed bug with Data List reports – wouldn’t update display after changing sort order of a variable.
- Sort by value now works properly in Data List reports.
- Can now handle excessively long values being used as categories in report tables or charts etc. Checks are also now made for excessively long category variable values.
- If encoding problems, SOFA now tries to use the field encoding e.g. iso-8859-1.
SOFA creates HTML web reports ready to share but sometimes you want to share a PDF or include graphs and tables in documents and slideshows. The ‘export output’ extension makes it easy. And the images can be exported in draft for speed or a range of higher quality resolutions e.g. for publication.
The Export Output plug-in is available at get_extensions.php and is compatible with SOFA 1.2.0, which has just been released.
Version 1.2.0 also offers the following:
- Scatterplots can have more than one series e.g. different coloured male and female dots on same plot.
- Individual columns within a data display report table can be sorted. This is extremely useful for checking your raw data, looking for anomalies before analysis.
- Report tables of summary statistics e.g. mean, median, sd, now show the measures across columns instead of down rows. This makes it much easier to compare results and notice difference.
Numerous changes to layout and functioning to make SOFA easier to use. For example, the demonstration tables displayed when configuring report tables now use real data instead of random data if the data source is small enough to make updating fast. And the “Show Results” button becomes “Add to Report” when report tables are able to use live data for examples (and the “Add to report checkbox” is hidden). Add to report now defaults to False and remembers the last setting throughout the user session. And some standard GUI items have been repositioned to increase the logic of placement and speed of use.
- Added IQR (Inter-Quartile Range) to the Row Stats Table.
- Can now do a boxplots of a single variable (not requiring a variable for it to be By).
- Now save and show cropped matplotlib charts.
- Scatterplots are a better size.
- Excel importing copes with faulty dates better and tells the user where the problem is.
- Filter dialog now checks if user wants to apply a filter when the settings haven’t been changed. Prevents inadvertent filtering to null.
- Better font sizes in auxiliary clustered bar charts (matplotlib).
- GUIs no longer update if the user selects the same database as is already selected.
And there have been plenty of useful bug fixes too:
- Fixed bug when more colours are required for auxiliary clustered barcharts than in the css.
- Fixed bug re: linked images with non-English characters not displaying in GUI html window (still issue if user relying on IE6).
- Fixed bug which made report tables widen if the title or subtitle was wider.
- Fixed bug where removal of filter in the charting dialog would change third variable dropdown to the last value it had when the filter was applied even if it had subsequently been changed to the Nothing Selected value. The default values are now updated everytime the dropdowns change.
- Now a guaranteed consistency in series colours across charts even if a series is missing e.g. the middle one, from one of the charts.
- Misc bug fixes for stats tests where no variability. Better user messages.
- Independent variable tests cope with new lines in text values.
- Histograms all get title case titles if no value label instead of upper case.
- Fixed bug when filtering results in no group a and b values possible in indep2var analyses.
- Fixed histogram width issue esp if wide numbers e.g. 12000 is wider than 12.
- Fixed css display problem in webkit internal browsers.
This is an exciting release and hopefully the plug-in will be well received.
I was wanting to shrink the font of elements of the SOFA GUI dialogs so I could squeeze more in or relocate items to more logical positions. Can’t be that hard, surely? I have since discovered that if a drop-down list (wxPython wx.Choice widget) has lots of items e.g. 30+ it takes seconds for fresh items to be added the the widget if you are trying to use your own font selection (using setFont()) on Linux. SetItems() takes a long time as, presumably, it sets the font for each individual item. And given I can’t control how many items will appear in drop-down lists or avoid having to repopulate lists (e.g. new data table selected so variable lists have to be updated) the option of shrinking fonts is not viable. Back to the drawing board.
[UPDATE] I came up with a workaround. Because there is no performance problem when items are included with the initial instantiation of dropdown widgets, all dropdowns are rebuilt each time they are changed. This means they have to be destroyed before being replaced, and the panel they are on must be hidden temporarily to avoid flicker on Windows, but it works. The fact that I was able to clean up some code in the process almost compensates for the considerable extra work
And when a data type mismatch is encountered during an import, SOFA now reports on the first faulty cell (row, cell value, and expected type) to make it easier for to clean your data.
In addition, here are the other features added in this release:
- Now able to run multiple single line charts. And multiple line charts have a consistent y-axis for role as trellis charts.
- SOFA now guesses whether an xls or ods spreadsheet has a header row or not when importing and sets the default buttons accordingly.
- Tighter and more flexible spacing of x-axis title according to number of lines in x labels. Now show legend title before legend and only show legend if multiple series.
- Smarter column resizing.
- CSV importing correctly identifies if file has a header more accurately if only one column (and thus no delimiter identified initially).
- Adjusted chart sizings to be more rational and less magic.
- Better width settings for dojo histogram output when lots of bins (prevents x-axis labels bunching up).
- Row stats tables show “Not calc” instead of nan (not a number) when appropriate.
And there are some important bug fixes as well:
- Fixed bug encountered when importing spaces in fields.
- Fixed bug when missing values in some fields in a boxplot with multiple series.
- Table design changes now cope with pre-existing sofa id index and demo values no longer updated on exit (inadvertently via on_show event).
- Fixed bug in tooltip display of averages in general charts – no longer rounds to lower integer.
- Charts of averages don’t show percentages in tooltips.
- All matplotlib histograms, including those in stats display such as for one-way ANOVAs, now cope with sigma of 0.
- Fixed bug calculating percentages where all values in a chart (of a series) are 0.
- Fixed bug in showing rotate when changing to clustered bar charts.
- Fixed bug in ods importing where display of sample values didn’t include repeated cells (only showing the first value).
- SQLite now correctly identifies more types as numeric.
I hope you like it.
More steady improvements to SOFA:
- Added ability to chart percentages as well as frequencies.
- Chart series now all have consistent y-axes (trellis style).
And there have been some bug fixes:
- Fixed ODS import bugs when encountering fields formatted as fractions, boolean, percentage etc.
- Fixed bug stopping linked images external to the html e.g. generated by matplotlib, from displaying in the internal GUI if the report path was different from the default report path.
- Fixed bug when running scatterplots and histograms with chart by (because of use of dd object which cannot be used headless).
- Fixed bug in charts dialog where second variable drop down would sometimes be overly restrictive when doing a chart of average values.
- Fixed bug where old pycs can interfere with Windows upgrades.
Charts now have the option of rotated (vertical) x-axis labels. This can be useful for longer labels.
Note – if you have upgraded SOFA rotated labels may not work unless you update the sofastats_charts.js file in your local sofastats folder e.g. C:\Users\username\sofastats\reports\sofastats_report_extras with the sofastats_charts file for sofastats_report_extras
Scatterplots now focus on the data better by starting axes just below the minimum x and y values of the data unless the value is close enough to 0 to make it worth using 0 anyway.
And for Ubuntu users, a much nicer launcher icon :-). Actually, it’s a set of icons at different resolutions so that SOFA always looks good on the launcher.
Other changes include:
- Numeric values are right justified in data tables.
- Kurtosis values in the normality test include the Fischer adjustment (subtracting 3).
- Duplicated field names in imports are given unique suffixes and allowed (now that they are unique).
- Excel importing now handles times without dates.
- More date formats are accepted when importing data.
- Better guidance on data preparation before importing data.
- More robust handling of variable definition files if corrupted.
Note: if upgrading on Linux, the two user folders (sofastats and sofastats_recovery) may be shifted from inside your home folder to a better location e.g. “/home/username/Documents” if free desktop standards are supported. After upgrading you may wish to manually replace the contents of the new folders with the contents of the old ones.
- Fixed small bug stopping column labels displaying in data table view.
- Fixed bug in recode operation which would wipe the table if any errors at all where encountered trying to turn the user recode config into SQLite update clauses.
- Fixed bug in getting structured data e.g. for line charts, where a user names a field freq and thus has a conflict with my own freq field. Renamed the internal use field _sofa_freq to prevent collisions.
- Creating user’s default proj file now copes with apostrophes etc in user path e.g. /Users/Jim’s/etc.
- The project dialog now displays the default report and css details saved with it from previous occasions.
- Project settings are only applied if the project is selected – they are not automatically triggered by changes when configuring a project.
- Multi-line values entered into data cells e.g. variable label settings, automatically have the line breaks converted into spaces. Prevents errors in display of data e.g. in single line text boxes, and problems storing in python scripts (EOL error) etc.
- Fixed bug where the first SQLite database in a project was assumed to be the default sofa database even though it might not be. Now possible to link to multiple default databases e.g. testing copies etc as long as simple naming convention followed.
- SOFA now rolls back to last good database connection if a failure.
- Fixed strange bug where default database would lock if made a new table, then looked as design, then tried to write to the database e.g. importing, editing data. Just refreshed cursor after updating demo table design and problem gone.