Map Stat Data
Map Stat Data
A. Original Data
B. Recalculated Values
54
Hands On
A. Counts: Original Data
G
N
O
R
W
C. Density: Original Data
G
N
O
R
W
D. Density: Recalculated Values
Figure 2: It is correct to use the chloropleth method to map densities but not counts.
55
interpreted as more), the map of recalculated values varies greatly from the map of
original values.
This violates the assumption that the
values in enumeration units are uniform
across the area. However, when density
is mapped, the distributions appear exactly the same. Arbitrarily dividing the units
does illustrate the properties of spatially
intensive and extensive data, but it is not
something you would probably ever need
or want to do.
Lets look at the actual data. The data
is not evenly distributed within the unit,
as is the case with most areal data. The
original units (shown in Figure 3) relate to
counties and are further subdivided into
census tracts. Mapping population density
for these tracts reveals that for the entire
area, the population is concentrated in
the southwest. However, mapping population by county masks this variation in
distribution.
There is also another problem with
56
mapping counts or totals and other spatially extensive data within areas using
the choropleth method. Distributions that
are uniform will be masked. The maps in
Figure 4A, 4B, and 4C show data mapped first
as a uniform distribution, then as two choropleth maps that display feature counts and
feature density. The count ranges widely by
area, causing a range of lightnesses on the
map in Figure 4B. Although the density is
the same for all areas, this variation gives a
false sense of the way the features are distributed within the areas. In contrast, the
map in Figure 3C has the same density for
each unit. The lack of variation in lightness
between units gives the correct impression
of feature distribution.
Figures 2 to 4 demonstrate a very important caveat: counts or totals and other
spatially extensive data should never be
symbolized using the choropleth mapping
method.
Why?
Because this method does not accurately
Normalizing or Standardizing
the Data
Now that you understand the need to
match the mapping method to the nature
of the data, the next step is to learn how to
work with the data so that it is in the correct
form for the type of mapping method you
are using.
Hands On
To correct the problems caused by mapping counts using the choropleth method,
you can convert the data to the correct
type so it can be shown by lightness within
areas. This is often necessary for data represented as points, lines, or rasters and
with other mapping methods such as proportional circle.
Do this by normalizing or standardizing the data. These two terms, often
used interchangeably, are slightly different. Normalizing the data scales all numerical values to a range from zero to one.
Standardization transforms the data so that
is has zero mean and unit variance. Both
techniques have drawbacks. If the dataset has outliers, normalizing will scale the
normal data to a very small interval. When
using standardization, the assumption is
that the data has been generated with a
certain mean and standard deviation, although this may not be the case.
Transformation
Operation
Ratio or rate = n a /n b
Proportion = n a /N
Percentage = n a /N * 100
Density = n a /A
A. Uniform Distribution
Figure 4: Uniform distributions (A) are masked using the choropleth mapping method to
show spatially extensive data, such as counts (B), rather than a statistic for spatially intensive
data, such as density (C).
57
A. Original Data
B. Density of Teachers
T: Number of Teachers
S: Number of Students
C. Percentage of Teachers
Figure 5: The original data includes counts of teachers and students. The area of the units was calculated with GIS. These spatially
extensive measures can be converted to spatially intensive data that can be appropriate to be mapped using the choropleth method.
Examples include densities (B), percentages (C), and ratios (D). (Data source: California Ed-Data website: www.ed-data.k12.ca.us)
58
Hands On
About the Author
are employed full-time. Two half-time
teachers may count as two teachers but
together are only one full-time equivalent
(FTE). This aspect of the data is not captured unless the number of FTEs is mapped
rather than the number of teachers.
Also, quantities that are not comparable
should not be used to calculate ratios. For
example, you would not calculate (or map)
the number of teachers per school unless
all the schools were roughly equal in size.
For this ratio to make sense, the schools
have to be comparable.
Summary
Understanding more about the nature
of the statistical data used for mapping
purposes will help you better understand
the methods that can be used to map it.
Ultimately, the goal is to match appropriate data with the most effective method so
that your map can be easily, quickly, and
correctly interpreted by your readers.
References
Brewer, Cynthia A. 2006. Basic Mapping
Principles for Visualizing Cancer Data
Using Geographic Information Systems
(GIS),
www.ajpmonline.org/article/
S0749-3797%2805%2900358-2/fulltext.
Cote, Paul. Effective Cartography: Mapping
with Quantitative Data, www.gsd.harvard.
edu/gis/manual/normalize/.
B
etter business analytics? Just add location.
Learn all about it at the Esri Business Summit.
July 79, 2013, San Diego, California
Register to attend at esri.com/aubizsummit.
59