Environmental Engineering Reference
In-Depth Information
Mean NDVI Rank
55 - 32,870
32,871 - 75,020
75,021 - 111,710
111,711 - 198,830
% African American
0 - 2
3 - 5
6 - 30
31 - 100
N
N
25 km
25 km
% High School Diploma
30 - 74
75 - 85
86 - 91
92 - 100
% Hispanic
0 - 1
2
3 - 4
5 - 89
N
N
25 km
25 km
FIGURE 16.1
Quantile-classified choropleth maps used in the analysis: Mean NDVI Rank (top left), percent African American
(top right), percent High School (bottom left), and percent Hispanic (bottom right).
in parts of Philadelphia (Fig. 16.1, top right), though there
are several smaller cities in the suburban and exurban parts
of the metropolitan area that also have high concentrations of
African Americans. Hispanics are also primarily concentrated
in parts of Philadelphia (Fig. 16.1, bottom right), though high
concentrations also occur in Chester County in the southwest
part of the study region, where demand for agricultural labor
draws Hispanic farm workers. High educational attainment is
concentrated primarily in the suburbs of Philadelphia (Fig. 16.1,
bottom left), with lower values in the exurban and rural areas
to the west and the lowest educational attainment evident in the
city of Philadelphia.
Results of theOLS regression of meanNDVI rank are reported
in Table 16.3. Note that poverty rate was dropped as an explana-
tory variable because it is highly correlated with percent high
school (
r
=
0
.
68, significant at
p<
0.005). The variance inflation
factor (VIF) statistic indicated problematic collineary among
explanatory variables when both poverty rate and percent high
school were both included in a single regression equation. All
regression models reported in Table 16.3 have VIF values less
than 2.0. Model 1 includes only the race variables in the model
and confirms the results of the Pearson correlations reported in
Table 16.2 - high NDVI is associated with low concentrations of
both African Americans and Hispanics. Using the race variables
alone, 25% of the variation in mean NDVI rank is explained.
Model 2 also includes percent high school, and indicates that
high NDVI is associated with educational attainment. Notably,
the influence of educational attainment on NDVI is higher than
that of either of the race variables, and reduces their influence
on NDVI from Model 1. When population density is added to
the regression equation, as in Model 3, both race variables lose
significance. The adjusted
R
2
also increases to 0.57.
The Moran's
I
statistic indicates that all the models pre-
sented in Table 16.3 have spatial dependency in the model
residuals - differences between observed and predicted values
of mean NDVI rank tend to be similar for tracts nearby one
another as compared to tracts farther apart. We employ spatial
econometric modeling estimation to account for spatial autocor-
relation in the model residuals. The spatial lag form of the spatial
econometric model incorporates the spatial lag of the dependent
Search WWH ::
Custom Search