Graphics Reference
In-Depth Information
frame of the original plot remains available and the subset is clearly visible as well.
Juxtaposition iswellknown forstatic plots buthasnotyetgainedacceptance ininter-
active graphical systems. here seems to be a psychological barrier that prevents the
use of juxtaposition because usually it requires rearrangement of the original plot to
makespacefortheadditional plot.Italso means that eachplothastobefullyredrawn
ater each user interaction, which can be quite demanding depending on the size of
the dataset and the complexity of the graphical display. But current computer power
shouldrendertheinteractive creation ofjuxtaposed plotspossible.Juxtaposition also
opens the possibility of seeing a full sequence of user interactions and thus allows
one to meticulously inspect the flow of subsetting. Juxtaposition also is a worthwhile
representation scheme for dynamic animations. he principle of juxtaposition can
be straightforwardly extended from graphical displays to statistical model displays.
Having calculated a statistical modelfor the whole sample, we might be interested in
particular subsets. Interactively specifying a subset of the sample population should
then run the model for the selected subgroup only. he new results could then be
juxtaposed to the original ones to allow for easy model comparison.
Special Forms of Linked Highlighting
8.3.4
Different problems occur when the linking scheme is not a simple -to- linking but
a more complex form such as m-to- linking as occurs in hierarchical linking. As-
sume that there are two levels in the hierarchy: the aggregated macro level, a set of
counties for example, and the micro level, a set of towns in these counties. When-
ever some but not all towns in a county are selected, it would be nice to represent
this partial selection of the county by a partial highlighting. If the representation of
the macro level is based on regular shapes, the partial highlighting can be done by
subdividing this shape into a selected and a nonselected component to indicate the
amount of selected objects on the micro level. he more general approach, however,
is to use different intensities of the filling color of graphical elements to represent the
various selected proportions. his is not only recommended fordisplays with graph-
ical elements that have a nonrectangular layout, but it is the more general approach
that isusually easier todecode.Figure . referstotheBavaria dataset ofFig. . and
showstwomaps,theletmapportrayingthemicrolevelof counties, therightmap
showing the macro level of regions in Bavaria. he selection of some elements of
themicrolevelintheletmapispropagatedtotherightplotandportrayedthere.he
varying intensities of the filling color reflect the proportion of highlighted counties
in a region.
Although the general linking schemes in this paper have been introduced under
the restriction of a directed linking process, a few comments shall be made on bidi-
rectional linking, which allows the mutual exchange and sharing of information be-
tween plots. Such bidirectional links would be useful when linking plot parameters
that govern the general layout and size of a display. In the unidirectional case, one
plot inherits the axis limits from the other. hese boundaries might be too small for
the passive plot and hence lead to a misrepresentation in the linked plot. It would be
much better in such instances to change both sets of plot parameters in such a way
Search WWH ::




Custom Search