Graphics Reference
In-Depth Information
variable production. Data are available for of the US states only. An identical
color links all of the descriptors for a region. Successive perceptual groups use the
same set of distinct colors. In Fig. . , the sorting is done (from largest to smallest)
by soybean yield in those US states where soybeans were planted. Here, the points
within a panel are connected to guide the viewer's eyes and not to imply that inter-
polation is permitted. he connecting lines are a design option and can be omitted
to avoid controversy or suit personal preferences. he list of US states is not evenly
divisible by five. Two perceptual groups at the top and two groups at the bottom con-
tain four states, while three perceptual groups in the middle contain five states. he
middle groups require the use of a fith linking color. Using distinct hues for linking
colors works best in full-color plots. For grey-level plots, colors need to be distinct
in terms of grey level. Figure . shows different shades of green and is suitable for
production as a grey-level plot. Readers new to LM plots sometimes try to compare
regions with the same color across the different perceptual groups, but they quickly
learn the linkage is meaningful only within a perceptual group.
While the micromaps in Fig. . are not ideally suited to show geospatial patterns,
the statistical panels nevertheless are more revealing than the choropleth maps in
Fig. . . It is immediately obvious from this graphical display that there is a positive
correlation between the two variables yield and acreage (high values for yield are as-
sociated with high values for acreage while low values for yield are associated with
low values for acreage). However, there were some considerable spatial outliers that
could not be detected easily in the two choropleth maps in Fig. . . Wisconsin (in
a favorable geographic region for soybeans, neighboring Iowa, Illinois, and Indiana)
had a high yield, but only a very small acreage was used for soybeans. On the other
hand, Arkansas, with a relatively low yield (similar to its neighboring states in the
Central South), used a surprisingly high acreage for soybeans. Geographically, the
highest yields for soybeans were obtained in the Midwest, with a spatially declin-
ing trend toward the Southeast. When visually comparing acreage and production,
these two variables almost perfectly correlate. Overall, the LM plots in Fig. . pro-
vide a much better and simpler geographic reference to the underlying data than the
two choropleth maps in Fig. . .
Design Issues and Variations
on Static Micromaps
1.3
Figure . shows one possible partitioning of regions into perceptual groups of size
five or less. However, for different numbers of regions there may exist more than just
one meaningful partitioning. Table . shows different ways to partition regions into
groups of size five or less. In fact, the partitioning in Fig. . is called Partitioning
in Table . . An algorithm, coded in S-Plus (Becker et al., ), produces symmetry
about the middle panel or the pair of middle panels (Partitioning ). It puts small
counts in the middle. For complete data situations, a single region appearing in the
Search WWH ::




Custom Search