Graphics Reference
In-Depth Information
Figure
.
.
[his figure also appears in the color insert.] Rug plot displaying the price evolution
(y-axis) of
online auctions over calendar time (x-axis) during a three-month period. he colored
lines show the price path of each auction, with color indicating auction length (yellow,threedays;blue,
five days; green,sevendays;red,tendays).hedot at the end of each line indicates the final price of the
auction. he black line represents the average of the daily closing price, and the gray band is the
interquartile range
the curve) and via color (different colors for different auction durations). Notice that
the plot scales well for a large number of auctions, but it is limited in the number of
attributes that can be coupled within the visualization.
Finally, trellis displays (Cleveland et al.,
) are another method that supports
the visualization of relationships between functional data and an attribute of inter-
est. his is achieved by displaying a series of panels where the functional objects are
displayedat different levels (orcategories) ofthe attribute ofinterest (see forinstance
Shmueli and Jank,
). In general, while static graphs can capture some of the re-
lationships between time series and cross-sectional information, they become less
and less insightful as the dimensionality and complexity of the data increase. One
of the reasons for this is that they have to accomplish meaningful visualizations at
several data levels: relationships within cross-sectional data (e.g., find relationships
between the opening bid and a seller's rating) and within time-series data (e.g., find
an association between the bid magnitudes, which is a sequence over time, and the
number of bids, which is yet another sequence over time). To complicate matters,
these graphs also need to portrayrelationships across the different data types; forex-
ample, between the opening bid and the bid magnitudes. In short, the graphs have
