Quantifying Urban Diversity: Multiple Spatial Measures of Physical, Social, and Economic Characteristics - Computational Approaches for Urban Environments - page 167

An examination of Fig. 7.11 shows that there are some well-defined areas of
maximum livability within Washington, D.C. The area between Dupont Circle and
Shaw/Mt. Vernon, extending north to Columbia Heights, as well as the general area
of Capitol Hill clearly appear to be the most livable areas within the city across all
four parameters. The UL index indicates where all four parameters are performing
well. The two high-performing areas in the composite index consistently perform
well across parameters. In other areas of the city, such as Georgetown and Historic
Anacostia, the UL index shows them performing at a mixed level.
In order to better understand the locational relationships of each of the parameters
and the UL index, an analysis of spatial autocorrelation was performed. First,
in order to confirm the presence of spatial autocorrelation (clustering), a general
Moran's I statistic was calculated for each of the parameters and the UL index, using
a distance threshold of just over one mile. While subjective, this distance can be
considered a reasonable distance to approximate the area that might be considered a
neighborhood for a resident of the subject block group, whether walking, bicycling,
driving, or taking transit for the mode of travel. Table 7.3 presents the Moran's I
statistic and associated z -score from each of the parameters and the UL index.
From the Moran's I statistic, it is confirmed that each of the parameters and the
UL index exhibit significant spatial clustering, as Table 7.3 shows both I and the
associated z -score for all parameters are greater than 0. However, this fails to inform
as to whether high or low values are clustering together.
For this final step in the cluster analysis, the Getis-Ord Gi* statistic was examined
for each parameter and the UL index, again, using a threshold of just over a mile.
Figure 7.12 a presents an unclassed choropleth map of the results of the Getis-Ord
Gi* calculation for the dwelling density parameter. From this map, it becomes clear
that dwelling density exhibits strong clustering of both high and low values, with
high values indicating higher densities, and low values indicating lower densities.
Figure 7.12 b is an unclassed choropleth map of the Getis-Ord Gi* calculations for
the mean street segment length parameter. This map depicts a strong clustering of
low values and a slightly weaker clustering of high values. Here, low values indicate
shorter mean street segment length, while higher values indicate longer mean street
segment lengths. Figure 7.12 c is an unclassed choropleth map of the Getis-Ord
Gi* calculation for the building year difference parameter. This map, again, shows
clear clustering of high and low values. Here, high values indicate a greater range
of building ages while lower values indicate a narrower range of building ages.
Figure 7.12 d is an unclassed choropleth map of the Getis-Ord Gi* calculations for
the mixed-use parameter (again, using the preferred mean shortest distance method).

Tabl e 7. 3 Moran's I and z-scores for all parameters and the UL index

Mean street
segment length

Building year
difference Dwelling density Mixed use UL index

Moran's I 0.44 0.25 0.18 0.28 0.4

Z -score 36.55 20.45 15.38 23.2 33.18