Graphics Reference
In-Depth Information
Visual Exploration by Linked Views
8.1
he basic problem in visualization still is the physical limitation of the
-D presen-
tation space of paper and computer screens. here are basically four approaches to
addressing this problem and to overcoming the restrictions of two-dimensionality:
. Create a virtual reality environment or a pseudo-
-D environment by rotation
that is capable of portraying higher-dimensional data at least in a
-D setting.
. Project high-dimensional data onto a
-D coordinate system by using a data-
reductionmethodsuchasprincipalcomponentanalysis,projectionpursuit,mul-
tidimensional scaling, or correspondence analysis.
. Useanonorthogonal coordinatesystemsuchasparallelcoordinateswhichisless
restricted by the two-dimensionality of paper.
. Link low-dimensional displays.
he idea of linked views has been around for quite some time in order to escape the
limitations of
-D paper or, as Tute (
) puts it, “the
-D poverty of endless flat-
lands of paper and computer screen.” Identical plot symbols and colors are a com-
mon way to indicate that different displays refer to identical cases. his has been
widely used in the development of static displays; see Tute (
) and Diaconis and
Friedman (
).InMcDonald (
)this concept of linked graphics wasfirstimple-
mented in a computer program to connect observations from two scatterplots. Still
bynow,thelinkinginscatterplotsandscatterplotmatrices,alsoknownas“scatterplot
brushing”aspromotedbyBeckeretal.(
),isthemostprominentcaseoflinked
views.
he main advantages of linked views are the easiness of the underlying graphical
displays and the speed and flexibility with which different aspects of the data can be
portrayed - three features that are essential in the exploratory stage of data analysis.
Linking a barchart with a histogram, for example, provides the opportunity to com-
parenot only those groupsthat aredefined byeachparticular category butalso those
that originate from uniting similar categories without actually changing the under-
lying data. Figure
.
shows in the background an average shited histogram for the
total population and in the foreground the average shited histogram for a selected
Figure
.
.
Does students reading behavior have an impact on their performance in mathematics? he
distribution for those students who do not read at all outside school falls below the overall distribution
