Graphics Reference
In-Depth Information
Rectangle or rubber-band selection. Forthisselectionmechanism,theuserclicksand
drags with the mouse to define a rectangle (or other shape, such as a circle) and,
on releasing the mouse, the graphic items intersecting the shape are considered
selected and their associated cases' degree of interest set to
.
Lassoing. Forthismechanism,the userclicksanddragstocreateapolygonselection
shape, either byclicking and dragging to define a shape orby successive clicksto
define a polygon. When completed, either byreleasing the mouse or bycomplet-
ing the polygon, the graphic items intersecting the lasso/polygon are considered
selected and their associated cases' degree of interest set to
.
To fulfill requirement (
), a view must be capable of displaying the base data for the
viewtogether with arepresentation of thedegree of interest. Ifwe think ofthe degree
of interest as a variable, then we can imagine it being used in the same way as any
other variable. Using the language of the Grammar of Graphics (Wilkinson,
),
we can use a variable such as:
A positional variable, used to define the geometry of the view;
An aesthetic variable, used to modify the appearance of a graphic element;
A faceting variable, used to define a paneling of the view.
In Fig.
.
below, barcharts showing the three basic methods are displayed. Note that
for the
-D barchart, the degree of interest could easily have been a continuous vari-
able; the locations along the z-axis would have been well defined. For the aesthetic
version, we split each bar in two to show the relative proportion of selected and non-
selected data. his requires the degree of interest to be at worst a finite number of
Figure
.
.
hree basic methods for displaying a degree of interest. he basic chart is a barchart of
counts of players at different fielding positions. he let view adds a “z” position variable to the bar
chart, splitting the dataset along the z-axis. he center view sets the brightness ofbarsbythedegreeof
interest, with the “interesting” bars shown in black.heright view facets (panels)viewsbythedegreeof
interest (although there are two facets, we term this a single “view”). he data show players' fielding
positions in baseball, and we have selected those players with more putouts than assists. Clearly this is
highly correlated with the fielding position