Software for teaching High School Statistics
One can argue that there is a tradeoff in statistical software between
its
"ease of learning" and its "power". Typically, those packages
that
are easy to learn are menu-driven and point-and-click driven.
Some
even have features designed soley for pedagogical purposes. On
the
other hand, other packages are designed to provide a great number of
powerful
and fast statistical tools as well as data management capabilities.
Warning: this page is by no means intended to be comprehensive.
Merely
a reflection of my own experience.
High Scorers on the "Ease of Use" scale:
Fathom --
this is predominantly a teaching tool, but in recent years has been
expanded to included multiple regression and other useful tools.
Easy-to-use simulation is provided with animation,
should one care to slow the calculation to view the animation. (And
this is quite useful, maybe even necessary, when demonstrating a
simulation to a class.) This contains
many features that make this possibly the first-choice for classroom
demonstrations.
Runs on PC and Mac.
DataDesk -- this probably wins
on both the "Ease" and "Power" scales. This
offers a wide variety of tools, some fairly sophisticated, and is
fairly
easy to learn. Still, the learning curve is slightly steeper than
Fathom,
and the animation capabilities, while fairly extensive, are harder to
use
than Fathom's. Supports multiple regression and ANOVA.
Runs on Macs and PC.
DataDesk can also be bundled with ActivStats, in
which
case it becomes a fairly complete textbook/software package.
ActivStats
is a wonderful reference book/self-guided learning tool.
Tinkerplots -- I'm
not sure whether this is easy to use or not. The target audience
is middle school, and middle school students seem to find this easier
to use than adults, particularly adult statisticians. Tinkerplots
is designed for graphics and exploratory analysis. Built ontop of
a Fathom platform, it has a radically different design philosophy than
other software. Roughly, you could say that other packages are
designed assuming you want to create, say, a histogram, and then make
it easy for you to create a histogram. Tinkerplots is designed to
let students construct their own graphics, and then is designed for
teachers help students transition from their initial graphics to more
formal and, one hopes, precise representations. Runs on PC and
Mac.
High Scorers on the "Power" scale:
SAS -- a very popular package in the
professional world. Everything and
the kitchen sink. Command-driven; you type the right phrase to
invoke
the appropriate package, and it provides output. Powerful data
management
capabilities. Difficult to learn. (In fact, for all of the
packages
in this category, it's fair to say that people usually take classes
just
to learn these software packages. UCLA, for example, teaches
Stats
130B "Statistical Analysis with SAS" and also 130A "Statistical
Analysis
with STATA.") Doesn't run on Macs.
Splus or R -- These are both statistical
programming languages. Certain fundamental
"objects" and operations make it easier to "roll your own" analysis
routines.
Common routines (say, regression) are pre-programmed, and there
is
a large on-line library of routines that can be downloaded. Great
graphics
capabilities, and hands-on control. For people who prefer stick-shifts
to
automatic transmissions. (Actually, for people who prefer
shifting
the gears themselves, without the clutch.) Splus costs money, R
is
free and runs on the Mac (classic and OSX) as well as PC.
Stata -- Stata is more like SAS
than Splus. It runs on Macs and PCs and
is slightly easier to learn than SAS. Runs
on Macs and PC.
UCLA ATS supports some nice Stata teaching modules. Browse a little
on the ATS support pages. You'll find help for teaching
Stata and SAS, should you need it. Considerably less expensive
than SAS. Also includes an even less expensive student version.
Xlispstat -- xlispstat is very
much like Splus or R, except that it's LISP. If
you are a fan of LISP, you will be overjoyed. Otherwise, you might not
be
convinced until you have seen the graphics (even inter-active
graphics)
that can be produced. Still, it does mean a heavy dependence on
the
parentheses keys. Best of all: its free!
Middle-Ground
Arc -- Arc runs as a
"front-end" to xlispstat and as such is a user-friendly
(i.e. menu-driven) package and also gives you all of the (considerably
less
user-friendly) flexibility of xlispstat. ARC does one thing very,
very
well: linear regression. In fact, more of linear regression
than
you probably thought existed. It was developed to support a book:
Applied
Regression Including Computing and Graphics, R. Dennis Cook and Sanford
Weisberg,
1999 John Wiley & Sons. This is a wonderful book, but
somewhat
beyond the High School level. Still, you might enjoy it. The
software
is free and runs on PCs and Macs (but not OSX).
Minitab
-- I have no direct experience with this myself. However, I'm
told
it's fairly easy to learn and fairly comprehensive. It has been
around
for awhile and so has an established track record. It can also be
bundled
with ActivStats. Not sure whether it runs on Macs, though.
SAS also publishes JMP
, which is more user-friendly, I'm told. I have no experience with it.
Runs
on Macs (not OSX) and PC.
SPSS is used primarily by social scientists, although I have no
personal experience with it.
Forget about it
Excel: As Paul Velleman (developer of DataDesk) put it
once:
you can also use Excel as a word processor, but you don't,
do
you? A fine spreadsheet, a really lousy statistical package.
And,
yes, I do know about the statistical "tools" add on. Still, why
struggle
trying to turn your Excel bar graph into a histogram when other
packages
do it easily and quickly (and correctly the first time) and let you
change
the number of bins? Yes, I know your school already has Excel; so
if
you want to teach them Excel, then use it. If you want to teach
them
Statistics, use something else. I would also point out that Excel
is
not nearly as easy to use as even some of the more difficult to use
statistics
packages. This is often overlooked by Excel-as-statistics
proponents
because they already know how to use Excel. If you
already
know how to use Excel, then admittedly it is easy enough to learn how
to
operate the statistics add-ons. But if you don't, you will find
yourself
puzzled and perplexed and aggravated at having to rely on an annoyingly
chipper
animated robot for assistance.
Here is one
nice analysis of Excel's statistical shortcomings, and Paul
Velleman has a nice explanation on the Math Forums.