Graphics Reference
In-Depth Information
ItseemstobepossibletoadapttheS-PlusmicromapcodeforuseintheRsta-
tistical package (Ihaka and Gentleman,
), which can be freely obtained from
http://www.r-project.org/. Although no full implementation of a micromap library
in R exists at this point, the basic panel functions and simple examples of LM plots
have been converted from S-Plus to R by Anthony R. Olsen.
Figure
.
isbasedondata fromtheNCImicromapWebsiteandillustrates several
variations of LM plots, implemented via the S-Plus rlStateplot library. his example
uses a special layout for
regions that differs slightly from the two suggested par-
titionings in Table
.
and calls attention to the middle state ater sorting the states
based on one of the variables displayed. Figure
.
bases the sorting on white male
lung and bronchus cancer mortality rates during the years
to
. Vermont is
the middle (
th) state in the sorted order and thus has the median value. Instead
of showing eleven panels with micromaps as suggested by both partitionings in Ta-
ble
.
, the micromap panel layout shows the middle state as a sixth highlighted state
in the micromaps immediately above and belowthe median divider. hislayout calls
attention to symmetry and saves space by removing the need for an eleventh mi-
cromap. he US federal agencies produce so many graphics and tables for the
US
states plus Washington, DC that it is worthwhile to develop a special layout for rou-
tine use. Other situations such as producing LM plots for the major regions of other
countriesmaybenefitfromdevelopingspeciallayouts.Carr(
)consideredspecial
layouts for the counties of selected states and made first use of the S-Plus functions
assembled in the rlStateplot library. his library was written for students to produce
LMplotssuchasinFig.
.
.
he micromap construction in Fig.
.
introduces another useful design feature.
It accumulates states when moving up from the bottom panel or down from the top
panel toward the median divider. Ater highlighting five states in a micromap panel
with foreground colors and black outlines, this design continues to show the previ-
ously highlighted states with black outlines in panels closer to the median divider.
his black outline lits the states into the foreground, and the micromap progression
shows a sequence of foreground contours. A cluster of states from the Northwest and
much of the inland US emerges when moving up from the bottom panel. he mi-
cromaps immediately above and below the median divider are duals of each other
except for Vermont, which is in the foreground of both of these micromaps. Human
perception does not treat foreground and background equally, so the logical equiva-
lence of the two panels maynot be noticed. he micromap just above the median di-
vidercallsattentiontomostoftheUSborderstatesotherthanthoseintheNorthwest.
Full-color LM plots typically use distinct saturated hues to highlight the selected
regions in each panel. he nonhighlighted black-outlined regions are shown in a de-
saturated hue such as light yellow while the background regions are light grey with
white outlines. he grey-level representation of Fig.
.
makes it harder to distin-
guish between foreground (black-outlined) states with the lightest colors and the
near-white states highlighted in panels further from the median divider. However,
the reader can still easily look at one of the two micromaps by the median divider,
notewhetherornotaparticular stateisintheforeground,based ontheblackoutline,
and know which direction to scan in to find the panel where the state is highlighted.
