Graphics Reference
In-Depth Information
Figure
.
.
Degree of interest for aggregated points. he let view shows the mean degree of interest
mapped to the size aesthetic. he right view divides each circle into a selected area and an unselected
area
Youcanclearlyseetheappearance ofthedesignated hitterstarting in
.Seeing the
positionalsoappearintheNLinrecentyearswasatfirstasurprisefortheauthor.he
reason is that only in recent years have there been regular-season games between the
two existing leagues. Previously, two teams from different leagues would only meet
in the World Series. Now it is common for NL players to play AL players. When they
do so at an AL ballpark, they play under their rules and so must field a designated
hitter. hus we have statistics for designated hitters in the NL.
he Grammar of Graphics (Wilkinson,
) provides a useful list of basic ele-
ments that are used in graphics. All elements can be used in an augmented coordi-
natessystemwhereonecoordinateisusedforthedegreeofinterestvariable,andall
may be faceted by a categorical degree of interest variable. For our final choice - us-
ing an aesthetic - we have already discussed points and intervals (bars). Since lines
and paths are hard to split into parts, we will generally use a summary statistic for
them or draw two different elements superimposed, one for selected and one for un-
selected. Areas and polygons (maps, for example) will similarly require a summary
operation. Areas have the additional advantage of being able to be displayed with se-
lectedandunselected subsetsstacked on topofeachother inasimilar fashiontobars
(e.g., Fig.
.
).
he schema element - defined in the Grammar of Graphics as a composite elem-
ent - is trickier. Examples of schemas are boxplots and Chernoff faces. For these ele-
ments, superimposition is hard to interpret; the elements generally look too confus-
ingly overlaid. A better choiceis toadd adimension todifferentiate between selected
and unselected subsets, as in Fig.
.
below.
Figure
.
issimilar toFig.
.
,withtheboxplotsreplacingtheparallelcoordinates
plot. It is much easier to compare the performance of highly paid players; they ap-
pear not to be paid for average batting and getting on base, but for power hitting -
homeruns and slugging. helinked boxplots allow ustocompare distributions more
