Should I use PSPP or R?

[Thomas Levine]

Hi,

It seems like the only real free statistics packages are PSPP and R, but I'm
having trouble understanding why people want to create a whole new system
even though R already exists. Aside from SPSS compatibility, what
advantages does PSPP have over R?

Thomas Levine

[Dr Eberhard Lisse]

SPSS compatibility of course.

So you can run your legacy programs, legally :-)-O.

I personally would like to have something like RPSS, ie an R module that
can not only read SPSS tables, but also SPSS programs, but I don't see
it coming just now, but, of course, you are most welcome to write it :-)-O

el

Post by Thomas Levine
It seems like the only real free statistics packages are PSPP and R, but
I'm having trouble understanding why people want to create a whole new
system even though R already exists. Aside from SPSS compatibility,
what advantages does PSPP have over R?

[Thomas Levine]

I haven't really used PSPP or R yet, but I don't understand how SPSS
compatibility was a reason to write a complete new system rather than a
module like the one were discussing. Can anyone explain this in simple
terms?

Post by Dr Eberhard Lisse
SPSS compatibility of course.
So you can run your legacy programs, legally :-)-O.
I personally would like to have something like RPSS, ie an R module that
can not only read SPSS tables, but also SPSS programs, but I don't see
it coming just now, but, of course, you are most welcome to write it :-)-O
el

Post by Thomas Levine
It seems like the only real free statistics packages are PSPP and R, but
I'm having trouble understanding why people want to create a whole new
system even though R already exists. Aside from SPSS compatibility,
what advantages does PSPP have over R?

--
Dr. Eberhard W. Lisse \ / Obstetrician & Gynaecologist (Saar)
PO Box 8421 \ / Please do NOT email to this address
Bachbrecht, Namibia ;____/ if it is DNS related in ANY way

[Ben Pfaff]

Post by Thomas Levine
I haven't really used PSPP or R yet, but I don't understand how
SPSS compatibility was a reason to write a complete new system
rather than a module like the one were discussing. Can anyone
explain this in simple terms?

R and PSPP are about the same age.

[John Darrington]

To answer your original question, if you've never used a stats
package before, then I would suggest that you start with R. Most
users are of the opinion that R is more powerful but harder to learn.

The primary motivation to develop PSPP (for me anyway) was to benefit
those people who have been using SPSS for decades and would like to
switch to Free Software but can't, or don't want to, learn a new system.

There are also other benefits of PSPP over R, including:

* Possibly faster.
* Copyright is held by one party only.
* (in the future) integration with desktop software.

An R module to read PSPP syntax might be a good idea, and the PSPP
code is now mature enough that it may be possible (if there's any
interested R developers out there, then talk to us). But parsing the
syntax and interpreting it is not trivial. The effort involved in
writing such a module from scratch, I suggest would be similar to
writing PSPP all over again.

J'

[Ben Pfaff]

Post by John Darrington
* Possibly faster.
* Copyright is held by one party only.
* (in the future) integration with desktop software.

I'm increasingly optimistic that we can develop a really
high-quality, user-friendly statistical analysis tool with high
performance, the capacity to efficiently handle large data sets,
high-quality human- and machine-readable output, and that is easy
to hack on too. It's getting better all the time. Let's keep
plugging away and see how good we can make it!

[Jason Stover]

Post by Thomas Levine
It seems like the only real free statistics packages are PSPP and R, but I'm
having trouble understanding why people want to create a whole new system
even though R already exists. Aside from SPSS compatibility, what
advantages does PSPP have over R?

A gui.

PSPP allows variable names with unlimited sizes. I don't know if R
does or not.

PSPP can handle pretty large data sets. R isn't designed to handle
large data sets, though it has gotten better at it over the years.

There may be some other concrete advantages that John and Ben
can mention.

In my opinion, at this point, other than compatibility and the gui,
PSPP does not yet have any clear advantages for most users. I use it
mostly because my students need to be able to use SPSS, and PSPP's
compatibility is therefore very important. Students in an introductory
Statistics course often just can't use R's syntax. They need a gui.

Other than compatibilty and the gui, the advantages of PSPP have more
to do with its "backend," which is something a casual user mostly will
not notice. We are trying to make its statistical functions more
modular than those in R. For example, PSPP uses GNU Scientific Library
for a lot of its computations, whereas R has its own implementation of
low-level functions.

-Jason

[Thomas Levine]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I haven't ever used any real stats package unless you want to count the
one on the TI-89 Titanium, and seems it won't really matter which one I
use as whatever I do should be pretty simple, so I'll use another
criterium that I sometimes use to determine which program to use: Is it
easier to convert files from PSPP to R or from R to PSPP?
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: http://firegpg.tuxfamily.org

iD8DBQFGP/ZWaROzpVxvjPURAklOAKCUrm45Ykre981bTPE40m1wB4onCACeLYgR
wI6tM3/fAMozcx/6g+urqEc=
=QFHe
-----END PGP SIGNATURE-----

Post by Thomas Levine
Hi,
It seems like the only real free statistics packages are PSPP and R,
but I'm having trouble understanding why people want to create a whole
new system even though R already exists. Aside from SPSS
compatibility, what advantages does PSPP have over R?
Thomas Levine

Thomas,
If you're happy with R, by all means. There are many people who use S/R
and love it... But there are legions of people who will never use R but
think SPSS is better than sliced bread.
IMHO it's pretty handy to have the spreadsheet-like interface to the
dataset when you are creating new variables, transforming variables,
etc. (Maybe R does this? I'm only familiar with the CL version. The
SAS version of this, where you can look, but not touch, is not as useful.)
For example, I recently had a dataset where I had 1-5 raters per person
and I wanted to calculate an ICC to assess agreement. To do this, I had
to take the dataset with one record per 1 rating and flatten it into a
dataset with one record per ratee with three ratings/columns (I
discarded the fourth and fifth ratings). SPSS provides functionality to
do this pretty neatly (and yet I screwed it up twice before getting it
right). Being able to write SPSS code to do the transformations but
also being able to visually inspect the data is a big plus for me.
(That said, R can do a lot of neat tricks as well.. especially
graphics... don't get me wrong.)
Ultimately, why would you only want one tool in your toolbox?
-Alan
--
Home computers are being called upon to perform many new functions,
including the consumption of homework formerly eaten by the dog.
--Doug Larson
Alan D. Mead, Ph.D. | Institute of Psychology
Assistant Professor | Illinois Institute of Technology
Scientific Adviser, | 3101 South Dearborn, 2nd floor
Center for Research and Service | Chicago IL 60616
Skype: alandmead
+312.567.5933 (Campus)
+815.588.3846 (Home)
+312.567.3493 (Fax)

[Alan Mead]

Post by Thomas Levine
I haven't ever used any real stats package unless you want to count the
one on the TI-89 Titanium, and seems it won't really matter which one I
use as whatever I do should be pretty simple, so I'll use another
criterium that I sometimes use to determine which program to use: Is it
easier to convert files from PSPP to R or from R to PSPP?

R and PSPP are not that comparable, especially given the current state
of statistical routines in PSPP. If you want to analyze some data then
SPSS (and someday soon, PSPP) is a good route. If you want to learn a
neat statistical programming language which, incidentally will allow you
to analyze data at some point, start with R.

Again, my recommendation would be to learn both so you can pick the one
that is easier/better for a given task.
--
Home computers are being called upon to perform many new functions,
including the consumption of homework formerly eaten by the dog.
--Doug Larson

Alan D. Mead, Ph.D. | Institute of Psychology
Assistant Professor | Illinois Institute of Technology
Scientific Adviser, | 3101 South Dearborn, 2nd floor
Center for Research and Service | Chicago IL 60616

Skype: alandmead
+312.567.5933 (Campus)
+815.588.3846 (Home)
+312.567.3493 (Fax)

[Daniel E WILLIAMS]

I've found both useful on the job for different kinds of tasks. I use R
when I need matrix algebra, forecasting, survival analysis, neural
networks, or optimization. I use PSPP and SPSS to transform large
datasets, usually for further analysis in R.

Dan Williams
Forecasting, Research and Analysis Office
Department of Human Services
State of Oregon, USA
503 947 5354
Hi,

It seems like the only real free statistics packages are PSPP and R,
but I'm having trouble understanding why people want to create a whole
new system even though R already exists. Aside from SPSS compatibility,
what advantages does PSPP have over R?

Thomas Levine