The choice of which
statistical package to use in an introductory statistics or advanced
statistics course can be determined by a number of considerations:
Which statistics package is the instructor most
comfortable with?
Popularity of the statistics package
Goals of the intended student user - will the student
be doing more involved research and development, or will they be
engaging in intermittent cursory usage?
Ease of use - are there drop down menus? How
easy is the syntax/language to learn?,
Flexibility
Cost for the student
Will the student be using modern, advanced statistical
technologies, or will they be relying mostly on well known classical
methods?
How important is high quality, publication ready
graphics (both exploratory and classical)?
Availability of the software during course work, and
after the student leaves the academic institution
Is there an active, supportive community of users?
How available are documentation, tutorials, and books?
Are there statistics textbooks that cover software
usage along with theory?
These are only a few of the considerations involved in
selecting a statistics package for a statistics course. In this
article, we bring two data analysis/statistical systems to the attention
of educators: "S-Plus" (the commercial version of the "S" language) and
the public domain "R" (free version of the "S" language). We
discuss the cost and availability of S-Plus and R to the community of
UNT researchers, instructors, and students.
S-Plus
S-Plus
incorporates the
object-oriented language S, developed at
AT&T Bell Labs statistics research
group (Lucent Technologies). Marketed by
Insightful Corp., S-Plus fits
statistical models as "objects", making data analysis much more flexible
than the older, procedural language approach (e.g. SPSS, SAS). S-Plus
incorporates a highly useable graphical user interface (see
this online tutorial for examples), along with the capability of
script based processing. Additionally, S-Plus allows the user to
"interact" with data and graphics through a command line interface.
The figure below provides an example of the S-Plus GUI interface:
S-Plus has an active world-wide user community -
S-NEWS.
Additionally, Insightful Corp. provides online versions of all
S-Plus
documentation (this documentation is also installed locally upon
software installation). Students, instructors and researchers will
be glad to know that many
books and
tutorials have been published on the S-Plus system. Advanced
researchers should be excited about the continuing expansion of the
S-Plus system with the newest statistical technologies available.
Insightful Corp. provides numerous
"experimental" research libraries at no-charge for download.
Currently, these libraries include: S+CorrelatedData (mixed
effects generalized linear models), S+Best (B-Spline methods),
S+Resample (bootstrap library), S+Bayes (bayesian analysis), S+FDA
(functional data analysis). Many of the libraries utilize both a
"drop-down" GUI menu system and a command line interface approach.
One particular library that could be particularly useful to introductory
statistics instructors is the S+Resample library. A
current trend in statistics education is to use resampling methods
(e.g. bootstrap & permutation methods) to illustrate empirical sampling
distributions and non-parametric confidence intervals based on the
empirical sampling distribution. One notable example:
Tim Hesterberg and
co-authors have teamed up with the authors of the highly acclaimed
"Introduction to the Practice of Statistics, Fifth Edition" by
David Moore and George McCabe, to produce a
book chapter
that integrates the bootstrap into the statistics curriculum at an
elementary level. This book chapter utilizes the S+Resample
library to provide easy accessibility to resampling methods at an
introductory statistics level. Tim Hesterberg has also written
about using resampling and simulation
methods in teaching statistics. Researchers who are interested
in "data-mining" methodologies can use S-Plus in conjunction with
Insightful Corp.'s
"Insightful Miner" product to explore undetected patterns in massive
datasets. A quick search on Google search engine demonstrates that
S-Plus is a popular system for research and instruction (e.g. a search
on "S-Plus" returned 482,000 hits).
Pricing and Availability of S-Plus at the
University of North Texas
Students can purchase an "Academic" version of S-Plus at the UNT
University Bookstore for $25. This is a specially licensed copy of
S-Plus (for UNT campus) that expires one year after installation
(MicroSoft Windows version). This academic version has all the
features of S-Plus "Professional", except that it expires one year after
installation. Insightful Corp. also provides a "Student"
version of S-Plus that is freely available at
http://elms03.e-academy.com/splus/ This version of S-Plus is
free, and has full statistical functionality of the academic version,
but: 1) Has a 20,000 cell or 1,000 row limitation; 2) Is
only for educational use; 3) Expires after one year; 4) Has a large
download (more than 100 meg). Students register at the website,
download the software, and are given a license code that enables the
software. The "Student" version of S-Plus is an attractive
alternative to the "Academic" version of S-Plus for those instructors
teaching a "long distance" learning course where students are incapable
of purchasing S-Plus from the bookstore. For full-time faculty,
S-Plus can be obtained at no cost from the
Research and Statistical Support
Office (RSS) at UNT. S-Plus is gaining in popularity (it is
already a favorite amongst professional statisticians); S-Plus excels in
incorporating
modern statistical methodology while maintaining a large inventory
of classical statistical methodologies; There are many tutorials,
advanced methodology books, and introductory statistics textbooks
that incorporate S-Plus. S-Plus compares favorably on the all
software-choice considerations enumerated above. That is, S-Plus
can accommodate both novice users and heavily research oriented
practitioners of statistics.
R
R is an open-source initiative whose aim is to create and distribute
the same high quality, "cutting-edge" statistical technology that S-Plus
is known for (see the R homepage).
Quoting from the R homepage:
R is a language and environment for statistical computing and
graphics. It is a GNU
project which is similar to the S language and environment which
was developed at Bell Laboratories (formerly AT&T, now Lucent
Technologies) by John Chambers and colleagues. R can be considered as
a different implementation of S. There are some important differences,
but much code written for S runs unaltered under R.
R provides a wide variety of statistical (linear and nonlinear
modeling, classical statistical tests, time-series analysis,
classification, clustering, ...) and graphical techniques, and is
highly extensible. The S language is often the vehicle of choice for
research in statistical methodology, and R provides an Open Source
route to participation in that activity.
One of R's strengths is the ease with which well-designed
publication-quality plots can be produced, including mathematical
symbols and formulae where needed. Great care has been taken over the
defaults for the minor design choices in graphics, but the user
retains full control.
R is available as Free Software under the terms of the
Free Software Foundation's
GNU General Public License
in source code form. It compiles and runs on a wide variety of UNIX
platforms and similar systems (including FreeBSD and Linux), Windows
and MacOS.
As a free alternative to S-Plus, R cannot be beat. Available to
the R system are hundreds of user contributed libraries that cover large
areas of both classical and modern statistics (see
UNT's R
server help page on installed packages). While S-Plus excels
at providing advanced functionality through a menu system, R excels in
providing breadth in statistical functionality (e.g. our own
RSS R Server has
587 libraries installed). Much of this statistical functionality
is not duplicated for the S-Plus environment. Partly, this is a result
of the R system being an open-source project. Since the R source
code is available to developers of statistical technology, much
integration of R with existing statistical tools, databases, and
operating systems has occurred. The
"Omegahat" project being the
prime example of such efforts. From the Omegahat website:
Omega is a joint project with the goal of providing a variety of
open-source
software for statistical applications. The Omega project began in
July, 1998, with discussions among designers responsible for three
current statistical languages (S, R, and Lisp-Stat), with the idea of
working together on new directions with special emphasis on web-based
software, Java, the Java virtual machine, and distributed computing.
We encourage participation by anyone wanting to extend computing
capabilities in one of the existing languages, to those interested in
distributed or web-based statistical software, and to those interested
in the design of new statistical languages.
R's
integration with web servers should be of particular interest to
instructors who are interested in web-based statistics courses.
For a number of years now, I have been using a modified version of
Rcgi to create online, interactive tutorials for Benchmarks articles and
introductory statistics courses. Our RSS Matters column has
a number of examples of using R to create interactive tutorials:
robust statistics,
kernel density estimation,
false detection rate,
robust correlation,
bootstrap, too name a few. If, as an instructor, you are
concerned about the lack of a default drop-down menu system for R, some
efforts have gone toward
developing a GUI system for the R system. The most notable of
these efforts is John Fox's R Commander
(see our past Benchmarks articles on this GUI -
Article1;
Article 2;
Article 3 - these articles are somewhat dated). See the main R
Commander website for the most recent updates. R Commander uses
both a drop down menu system and a script window. Similar to other
statistical packages, R Commander pastes syntax into a syntax editor
whenever the contents of a menu system window have been submitted.
This allows easy access to default syntax (via a GUI) , but allows the
user to see the syntax, change the syntax, and save the syntax, for
later submission. This facilitates learning to program in the "S"
language. A couple of examples of R Commander's interface is presented
below:
Like the S-Plus user community, the R user community is highly active
as well - R-HELP.
In addition, the R developers publish a high quality, edited
newsletter that
covers software development news, R package development and usage, as
well as the usual tips and hints about using R. The user community
is also quite generous in providing
free tutorials,
books, and documents on R. R's
documentation is
very high quality as well. The basic R
language is well documented with examples that can be executed as is,
then modified as the user needs. For example, fitting a
regression, ANOVA, or ANCOVA model can be fit with the
"lm"
function. The help function for lm gives the user an
example that can be executed by pasting the text into the R console,
then altered as needed. The
"foreign" package gives users the ability to import other file
formats: SAS, SPSS, Stata, Minitab, SYSTAT, to mention some of the
more common formats available. R's base language is mostly
compatible with the S-Plus base language (greater than 95%?). That
is, most code written with the base R language will run unaltered in
S-Plus and vice-versa. It is not inconceivable that a student or
researcher would use both R and S-Plus in conjunction with one another.
Conclusion
In summary, R compares favorably with S-Plus (and is arguably
superior in some ways). In regards to some of the
statistical-software choices enumerated at the beginning of this
article: 1) Both S-Plus and R are readily available and
inexpensive to the student and instructor; 2) Both S-Plus and R are
readily available to instructor and student; 3) Both S-Plus and R
are inexpensive alternatives to more popular statistical packages (e.g.
SAS, SPSS, Stata); 4) Both S-Plus and R excel at providing a broad
range of classical and modern statistical methodologies; 5) S-Plus
utilizes an advanced menu system that is more accessible to students,
however, R is gaining some ground on that issue; 6) Both S-Plus
and R can accommodate a range of users from novice to advanced, that is,
both cursory users and researchers; 7) Both S-Plus and R have high
quality documentation and textbook usage; 8) The user communities
of both S-Plus and R are highly active and accessible to both student
and researcher; 9) S-Plus and R are already favorites
amongst theoretical and applied statisticians, and both of these systems
are becoming increasingly important in the environmental, biological,
medical, and social sciences, as evidenced by the increase in classes
being taught utilizing these environments and the increase in
statistical texts being published; 10) And most importantly - THE
PRICE IS RIGHT!
