Graphics Reference
In-Depth Information
angles, one for each case. Both drawing and brushing over the bars is handled as if
the bars were a collection of separate little boxes.
Oneofthemorepowerfulnovelfeatures ofLispStatisitsimplementation inLisp-
as an interpreted language, the user is free to write any function that can link views
together, and indeed can write any desired data view. If you want to interactively
colour items in a scatterplot based on their distance from the center of the brush (as
we do in Sect.
.
), it is an easy job for Lisp programmers.
Soforbarsonetechniqueistostacksectionsofthebarontopofeachother-either
grouping all the selected items together or leaving them in a fixed location on the
bar. hen we can use an aesthetic (typically colour or pattern) to colour the separate
sections. Another technique was alluded to in Sect.
; instead of dividing a bar up,
we use a summary function to create a single value for the bar that summarizes the
cases it contains and then map that to an aesthetic. his approach is more general
in that it can handle numeric degrees of interest beyond the
-
binary selection
needed for stacked bars, and it is also generalizable to graphical objects that cannot
be as conveniently split into components, such as lines and paths. Both techniques
are valuable and should be considered for use in any set of linked views. Figure
.
shows both techniques in action.
his figure combines five different variables for over
cases - aggregation is
clearly needed and a monolithic view that would adequately show the data is not ob-
vious.Inthebottomletpanelweshowahexagonal binningofthedataandrepresent
each bin with a circle. he mean summary statistic has been applied to the degree of
interest for the cases in the bin, each of which is either
or
. his effectively gives
us the fraction of the elements that are selected. Figure
.
below shows two alterna-
tive ways of displaying the information, using the size aesthetic and by dividing up
the glyph. In both cases the mean degree of interest maps to a circle area, not radius,
facilitating a better mental model of the distribution.
Although Fig.
.
is easier to interpret formost people,Fig.
.
might be preferred
in certain circumstances. For example, if we wished to use hue to encode another
characteristic, the mind has a tendency to combine hue and brightness into a gestalt
colour, which would make a colour aesthetic added to the scatterplot of Fig.
.
a more confusing view than if we added colour to either view of Fig.
.
. Conversely,
we might want to show the distribution of heights and weights rather than just the
fraction of designated hitters in each bin. To do that we might use a size aesthetic to
show the count of players in each bin and apply that aesthetic to the scatterplot of
Fig.
.
.
From Fig.
.
we can see that, although designated hitters tend to be of average
height,with relatively fewveryshortortall players,theyarewelloveraverage weight.
his makes intuitive sense if we consider their role. hey are players who do not
field or pitch. heir only role is to bat when the pitcher would otherwise be batting,
replacing him in the batting order. he designated hitter position was introduced
because the gap between the pitcher's ability to hit the ball and the rest of his team's
ability is generally large. herefore, rather than have a known poor performance on
every rotation through the team, one league, the American League (AL), decided
to introduce this position. It is a “power hitter” position. he designated hitter has
