Re: What good open source statistical software is available?

[Ben Pfaff]

irishhacker@gmail.com writes:

> On Apr 13, 11:31 pm, Ben Pfaff <b...@cs.stanford.edu> wrote:
>> "Robert" <callin...@yahoo.com> writes:
>> > What about these DAP and PSPP projects? Are they fairly mature and
>> > ready to use yet?
>>
>> I don't know whether I'd call PSPP mature, but we're actively
>> working on making it better all the time.  Your suggestions and
>> bug reports are welcomed.

> At the moment, I'm looking for work and living in a rented room.
> But when I get my own apartment ( hence my own home office), I'll
> download your product, your user's manual, and kick the tires.
> (i'm actually at an apple store now). Because I do want to see how far
> along
> your product is.

When you do get the chance, I'd recommend trying the most
snapshot on ftp://alpha.gnu.org/gnu/pspp instead of the most
recent released version (0.4.0).  The snapshot is (likely) better
quality and (definitely) has more features.

> I'm looking in your tarball for test programs(programs written in
> the , uh, PSPP
> programming language, with corresponding printout files, to see that
> the product was tested for that
> example and any bugs that popped up were resolved).

We don't provide many examples, but we do have a large number of
tests (137 in my working tree here).  We try to support the same
language as SPSS; thus, you can use any SPSS code you find
online, as long as the features that it uses are among those
supported by PSPP.

We would like to provide more examples.  If you do happen to
develop any nice examples of PSPP use, then I'd happily accept
contributions of them into the PSPP tarball.

> I'm a little confused by what I see, though. Some files seem to have
> some sample "SPSS" code, but mixed in with code
> in a very different syntax, and maybe the printout is in the same
> file(!), at the bottom of the file.

I think that you are probably looking at the tests.  These are
primarily written as Bourne shell scripts.  As you say, many of
them consist of a PSPP program followed by a set of expected
output to compare against the actual output.  Thus, if anything
changes in the output, we know that a bug has been introduced.

> Now SPSS and PSPP are recursive, right? So you can have an assignment
> statement, inside a while-loop, inside an if-else statement, etc. ,
> like
> you could do with any procedural language?

Well, to some extent, yes.  Some constructs--notably, statistical
procedures--can only occur at the top level.  And the semantics
of PSPP and SPSS are somewhat unusual, in that all the statements
(called, in PSPP and SPSS, "transformations") are within an
implicit loop over all the cases in the current data file (the
"active file").

> Also, I notice, looking at your Usenet postings, you never ever
> mention PSPP. Why?

It's rarely on-topic in the newsgroups I post in, it is rarely
relevant to the questions that folks discuss in those newsgroups,
and, finally, I'm not fond of tooting my own horn.

Hope this helps.

P.S. If you have more interest in PSPP, then you may want to join
the discussion in our users or developers lists, for which you
can find information at
        http://lists.gnu.org/mailman/listinfo/pspp-dev
        http://lists.gnu.org/mailman/listinfo/pspp-users
You can send mail to pspp-dev@gnu.org or pspp-users@gnu.org
directly, but if you don't join them first you'll have to wait
for someone to approve your first post, which can take a day or
so.
-- 
"I don't want to learn the constitution and the declaration of
 independence (marvelous poetry though it be) by heart, and worship the
 flag and believe that there is a god and the dollar is its prophet."
--Maarten Wiltink in the Monastery