Quantifying Urban Diversity: Multiple Spatial Measures of Physical, Social, and Economic Characteristics - Computational Approaches for Urban Environments - page 161

as a histogram and shows the rescaled values as an unclassed choropleth map.
Some of the greatest variability is located in the areas around Shaw/Mt. Vernon
Square as well as Historic Anacostia and Columbia Heights. These are all areas
where are currently undergoing significant new development and seeing noteworthy
demographic changes.
The area of the National Mall again stands out as something of an anomaly. This
is largely explained by the small number of address points within this block group,
thus having even one very new structure can easily alter the standard deviation for
this area.
For the fourth parameter, mixed primary uses, the results have been calculated for
the two methodologies: mean shortest distance between residential and commercial
locations and the distance between the mean center for residential locations and the
mean center for commercial locations. For the mean shortest distance, the values
calculated range between 18.1 and 3,047 ft with a mean of 499.2 ft and a standard
deviation of 482.1 ft.
A total of 94 block groups contained either no residential locations or no
commercial locations and thus were automatically assigned the highest rescaled
value of 1. Figure 7.6 shows the rescaled values as a histogram and the spatial
distribution of the rescaled values as an unclassed choropleth map.
For the second method, distance between residential mean center and commercial
mean center within a block group, the values ranged from 14.37 to 3,397.49 ft with
a mean of 555.51 and a standard deviation of 506.44. The same 94 block groups
with no mixed use were automatically assigned a value of 1. Figure 7.7 shows
the rescaled values as a histogram and the distribution of the rescaled values as
an unclassed choropleth map.
While both methods produce similarly shaped distributions, it is clear that the
results are spatially dispersed differently. The nature of the mean center method
means that it is less sensitive to localized conditions, making it a less suitable
method for measuring uses at a fine-grained level. Suppose, for example, that a block
group contained a cluster of residential units in the center and contained commercial
uses along its boundary; the mean center method would place mean centers for
both residential and commercial very close to one another, even though they are
not highly interspersed. The same method would produce similar results for a
block group that contained evenly distributed commercial and residential locations,
even though, by Jacobs' estimation, this second example contains a preferential
configuration. For this reason, the first method, the mean shortest distance between
residential and commercial locations is preferred. This first method is more sensitive
to clustering and dispersion within the block group.
As noted in the methodology section, it was thought that both of these measures
might be sensitive to the size of the block group for which they are calculated;
however, this proves not to be the case. Figure 7.8 is a scatterplot of the size of each
block group against its rescaled value for the mean shortest distance calculation.
Figure 7.9 is a scatterplot of the size of each block group against its rescaled value
for the mean center difference calculation. In both instances, there is only a very