Geoscience Reference
In-Depth Information
Table 3.4 Logistic regression analysis for patterns in consistency of ranking
Predictor
2
N
SE
b
Wald's
w
df p
e
b
(odds ratio)
b
Consistency
1220
% Latino
1.946 0.541
12.923
1
0.0001 0.174
<
% School-aged children
1.668 1.242
1.804
1
0.179
5.301
% Renters
4.387 0.468
87.881
1
0.0001 80.411
<
Constant
0.686 0.251
7.452
1
0.006
0.504
2
Cox and Snell R
2
Tests of model coefficients
p
w
Overall model evaluation
105.306
3
0.083
High information effort
1
ΒΌ
% school-aged children) and location near high or low levels of water information
exposure. In each of the analyses (Table
3.3
), high is the reference category.
All reported results indicate the increase in probability of a region's rank as an
area of high water information exposure with a unit increase in the predictor
variable. The Cox and Snell R
2
for the models ranges from
<
0.001 to 0.40
indicating a broad spectrum of model fits across regions (West Valley, Phoenix
and East Valley) and mapping techniques. Models using zip codes have the highest
adjusted R
2
values and results. Variables are significant most often in the distance
buffer models (see Cutts In Review for a full explanation).
In a logistic regression comparing the areas consistently ranked high to those
consistently ranked low, finding that % Latino was negatively significant, there was
no relationship between high water information and % school-aged children, and %
renters was positively related to water information availability. The full model
appears in Table
3.4
.
3.7 Discussion
The purpose of this study was to examine the extent to which effort to inform the
public about water issues might result in some populations having access to lower
amounts of information than others. Representing the data through seven separate
maps led to changes in the significance and direction of factors related to the
distribution of water information. Latino populations were less likely to live in
high effort regions (Prediction 1) in the West Valley and when we consider only
consistently ranked regions. However, Latino/a populations were significantly
more
likely to live in high education neighborhoods in Phoenix across two repre-
sentations and in the East Valley in on representation (Table
3.3
). When significant,
the proportion of the population renting their home was always
larger
than expected
by chance. This is counter to the expectation. Similarly, populations of school-aged
children are
smaller
than expected by chance, in direct opposition to out expectation
(Prediction 3). The observed differences within region as well as shifts in the
direction and significance of the models between the West Valley, Phoenix, and
the East Valley indicate support for geographically-based differences. While no one
