Software
There are many choices for statistical software at various
levels of technical skill and ease of use. The American
Statistical Association (ASA) Section on Statistical Computing
has put together the following listing: http://www.amstat.org/profession/index.cfm?fuseaction=software
- design. It describes features and benefits for some
common and specialized packages.
Graphics
What makes statistical packages distinct from business
software or commercial computer languages is the availability
of analytic graphics tools. These graphical methods are
not just for illustration. When graphs are designed with
appropriate methods, they let the viewer extract relationships
between levels, identify anomalies, simplify the complex,
and make better decisions.
Charts and graphs, which once required manual drawing
and then specialized programming, infuse all of daily life
now. In offices, Powerpoint® slides
filled with bar and pie charts are expected with any table
of numbers. These graphs are used in nearly all mass, electronic
and specialized media, most notoriously in the newspaper
USA Today. For all of the choices of color
and shape, there is little thought put into what the graph
encodes and what the viewer can glean from the flashy display.
In the past 25 years several thinkers have emerged to
address the graphics quandary. Edward
A. Tufte, Professor Emeritus of Yale, has published
4 ‘coffee-table’ books that explore how data
and images can connect. A true Renaissance man, he uses
examples from Gallileo, the Napoleonic wars, fish ecology,
aerodynamics, and contemporary sculpture. He is known for
coining the phrase ‘chartjunk’ and his recent
tirade against bullet-point style Powerpoint®
presentations. (Wired Magazine, September
2003). William
S. Cleveland, formerly of Lucent/Bell Labs and now a
professor at Purdue University, wrote two more pragmatic
books, The Elements of Graphing Data and Visualizing
Data. Since their publication, the concepts have spread
to other texts and software manuals.
Tufte and Cleveland advocate ‘small multiples’,
displaying many graphs of each level of a category on one
page or screen. Cleveland was involved in the development
of the S language at Bell Labs. This language has evolved
into the S-PLUS software package marketed by Insightful,
Corp. The S-Plus User’s Guide, available
on the Insightful website, takes this example of Barley
yield data from Visualizing Data.
Each site for the planting is a separate panel on the
trellis. There are 10 varieties labeled on the side of the
panels. The red and blue are yields for different years.
Trellis graphs can be used to show any kind
of category. A type of trellis is a subject plot. A subject
offer a visual story than can be hidden when only looking
at averages and statistical tests.
Spotfire,
goes even further, perhaps to the extreme.
The five graphs above in a tiled display (courtesy, SPOTFIRE,
Inc.) show the results of a drug screening for a cancer
compound that should inhibit tumor growth. Clockwise, the
tile combines a bubble plot, a three-D graph with response
by color, a trellised heat map that looks at 9 plates at
once, a segmented histogram, and a two-D plot with response
by color. The power in an active SPOTFIRE session is to
select out data simultaneously in the displays, dig deeper
into sections with the slider bars, and make discussing
large data more productive.
Books
Every year publishers hope that the new textbook will become
the definitive book. The book is definitive for you if it
helps you better understand concepts. Older texts often
are clearer with more details on statistical tests and concepts.
They serve as the basis for much of recent work. However
they can be obsolete regarding computation and graphics.
Contact us at 201-673-4301
if you want to discuss a book for yourself and your team.
