Graphics Reference
In-Depth Information
When we have to write programs for statistical data visualization using a general-
purpose programming language, for example, to achieve new functions or exten-
sibility, the Java language is an appropriate choice. he object-oriented character-
istics of Java are suitable for building graphical and interactive programs. Java has
well-prepared standard graphical libraries that reduce our programming tasks and
increase the portability of sotware to different platforms. Furthermore, recently de-
veloped solutions for object-oriented programming - so called “design patterns” -
are useful for building statistical data visualization programs.
his chapter illustrates howto build general-purpose extensible statistical data vi-
sualizationsotwareinJavabyfollowingdesignpatterns.Asanexample,weuseaJava
statistical graphics library named Jasplot (JAva Statistical PLOT).he source code of
Jasplot is available from our web site (http://jasp.ism.ac.jp/Jasplot/).
Introduction
17.1
Data visualization or statistical graphics is one of the important topics in statistics
(Wilkinson,
). Statistical graphics are usually drawn by computers these days.
hanks to the development of computer technologies, beautiful and colorful graph-
ics are easily drawn, and interactive and
-D graphics are even available on personal
computers (Symanzik,
). Almost all recently developed statistical sotware in-
cludes advanced statistical graphics functions, e.g., R (Murrell,
).
Although statistical sotware products have flexible and well-organized graphics
functions, we sometimes need new functions that have not been implemented and
are also di
cult to realize in suchsotware. For example, it is di
cult to realize com-
plicated interactive operations for graphics using only the traditional R language at
this stage. We note that the iPlot project (Urbanek and heus,
) expands the
capabilities of R to provide interactive statistical graphics using comprehensive Java
programs.
Before computers became available, it was di
cult to draw statistical graphics for
data that had many observations and variables. Now, such graphics are easily drawn
bycomputers. However, we know that the numbers ofobservations and variables be-
ing recorded have also increased with the development of automatic data collection
systems that use computers in various fields. Traditional statistical graphics some-
times fail to show the characteristics of such data clearly. For example, if we plot all
data on a screen using simple graphics such as a scatterplot or a parallel coordinate
plot (Inselberg,
), it is di
cult to distinguish each observation from the others
because of overlapping points. To solve problems like this, several interactive tech-
niques such as focusing, brushing, zooming and linked views have been developed
(Unwin et al.,
).hese techniques are useful for viewing particular observations
clearly and for grasping characteristics of the data structure.
It is convenient touse general-purpose object-oriented programming (OOP) lan-
guages such as C++ or Java to program statistical data visualization. OOP is a para-
digm where programs are composed by utilizing various “objects” in which related
