PSPPire problems

[Matej Kovacic]

Hi,

I am using Ubuntu 8.10 and PSPPire 0.6.0.

I am writing a review of the software for one of the biggest slovenian
IT websites, slo-tech.com (~250.000 unique visitors monthly). As
opensource fan, I am really excited about this software, but found some
bugs, which could be (in my opinion) quickly solved.

1. Problem with unicode support.
I found some problems when trying to open SAV data file which contains
data in unicode (slovenian characters čšž/ČŠŽ) and/or variable names in
unicode. Instead of slovenian characters I get just queston marks "?".


2. Importing data.
There is a lack of importing data from Excel, OOo Calc and dbase format.
It would also be fine to have support for importing data from Access
database and to have ability for direct connection to SQL database (for
instance MySQL and/or Postgres)


3. Exporting data.
Unfortunately there is no exporting to other formats except SAV and POR.
It would be fine to have at least exporting to tab-delimited TXT and
CSV, OOo Calc and Excel would be just great.


4. Merging datafiles.
Merging is very useful function and I would be very glad to have option
to merge datafiles (add cases and add variables). Aggregate function
would also be useful.


5. Count function in menu.
In Transform menu I would liek to see Count function. In fact in syntax
it is implemented, so it is probably trivial to implement it into psppire.


6. Selecting cases.
Selecting does not work similar to SPSS - in fact, just selecting cases
(not deleting them) is not working at all. SPSS creates some special
binary variable, which selects cases (1 or 0), and unselected cases are
crossed out.

There is also no option to select cases based on user-defined criteria
(IF (variable01 = 5) and variable02 < 8) --> select case)

Random sampling is also not working. I have a very large data set
(>500.000 cases) and tried to make a sample of 100 - nothing happened.


7. Output.
OK, for output you know exactly what should be done, probably better
than me. :-)


8. Additional analysis.
Do you have some priority list for implementation of statistical functons?
I am teaching at the university and working in public opinion research
institute and my recommendation would be: correlations, factor analysis,
clustering (especially dendrogram), principal component analysis.
Maybe you could also "copy" that functions from R-project, which is also
opensource?



However, the majority of problems is related to data manipulation and I
believe it is not really a problem, because data manipulation
(importing, exporting, etc.) is something which is already implemented
in various opensource programs and could be "copied".

Another set of problems in related to user interface (yes, graphs are
also something importand), but is already at least half-implemented (GNU
Plot, HTML/Postcript output in pspp).

The third set of problems is related to additional statistical analysis
methods. I really don't care if PSPP does not have all the SPSS methods
implemented, but if there will be implemented the most used ones, I
think our faculty and research institute can immediately "switch" to
PSPP opensource solution.

I understand this is 0.6.x version, but it is really great sotware and
with small improvement could became "SPSS killer" at least in some
segments of market.

P. S. - if you need test datasets with unicode data/variables, let me know.

Regards, Matej