New posts on the blog at
insideout.spss.com:
Python and Productivity: a story about
creating the SPSSINC TURF extension command.
A Lesson about Optimization: optimizing
SPSSINC TURF.
Stylin': Using templates for custom styles.
This includes a template download for styling mosaic plots.
New items on Developer Central (www.spss.com/devcentral)
A new extension command, SPSSINC CREATE
DUMMIES (the binary kind, that is). This extension command with
dialog interface creates a set of dummy variables corresponding to the values
of an input variable. It can also create a macro variable listing the
dummies. This can be useful, for example, when you want to use the
Regression procedure with a categorical predictor. It requires SPSS
Version 17 and the Python plug-in but can be made to work with version 16,
minus the dialog, with a few small changes listed in the readme file. No
Python knowledge is required to use it.
The SPSSINC TURF command has been updated.
As I mentioned in an earlier message, the TURF problem is computationally
explosive. Sure enough, people started using it in situations where it
had to calculate millions, billions, even trillions of set unions. It
didn't scale well for a reason I have written about on the blog. So part
of the code has been reworked to conserve memory and process the final results
faster. It still may take a long time with a big problem, but it now
scales much better than previously.
There is a small bugfix to the
ExportTablesToExcelFiles. download to handle the case where a table title has
line break characters in it.
The function genSortedVariableExpr has been
added to the spssaux2.py module. It creates a macro listing variable
names sorted by a statistic in a form suitable for a Ctables expression or an
ordinary variable list.
I will be speaking about SPSS and R at the
SPSS Higher Education Road Shows in Dallas on Thursday, 3/12 and Los Angeles on
Friday, 3/13. There may still be space available if you are in the area.
Jon K. Peck
SPSS Inc.
[hidden email]
(ip) phone 312-651-3435
Locked
