Thanks to Kit Baum, a new package -tableplot- has been added
to SSC. Stata 8 is required. To install,
. ssc inst tableplot
If this seems like d�j� lu, please note that -tableplot- is
not -tabplot-.
What -tabplot- does is
1 Calculate for a two-way table cell frequencies, fractions
or percents
2 Show them as a "table" of bars. It is just a wrapper for
-twoway rbar-.
What -tableplot- does is different.
1 You have to specify something to be plotted. That
must be something that takes just one value for each
cell of a two-way table. This has to exist as a variable
beforehand. It could be say a set of residuals from a model,
or a summary statistic of some response for the cells
of a two-factor table.
2 You must choose -rbar-, -rspike-, -rcap- or -rcapsym-
as a -twoway- plottype. (I miss an -rdropline- for what
have been called lollipop plots, in which the point
symbol on the end of a spike emphasises direction:
O
|
|
|
-------------------------
|
|
O
In the next version I may emulate that anyway.)
In these programs, I'm playing in public. Sometimes
I draw the graphs, and think "There is no gain over
a table here: in fact a table is much better!". Sometimes
the graphs make something more evident, or more vivid.
Of course, most graphs never make it to the printed page
anyway, and a useless graph is just a few seconds' work
discarded, once the program exists. But if anyone finds
these programs useful (-tabplot- or -tableplot-) I'd
appreciate a private note, with comment on what helped.
This should find a formal outlet in Stata Journal 4(2)
to appear in June. Stata Journal 4(1) includes
a piece on graphing distributions and should be out
next month.
This is also a partly a reaction to mosaic displays
or mosaic plots, which can be found in the literature
or on the web. There is an excellent introductory
tutorial at
http://www.stat.auckland.ac.nz/~ihaka/120/Lectures/lecture17.pdf
and mosaic plots are available in various other programs
(e.g. R, as the provenance above would lead you to expect).
It's a very neat idea, and attractive because they extend
to multi-way tables. A Stata implementation would be
most welcome. In practice they require more getting used
to than the rather conservative plots produced by -tabplot-
or -tableplot-, as you have to learn to focus not on the
area encoding but on lengths along one axis.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
