Title: Using Exploratory Spatial Data Analysis to Examine Neighborhood Structure and Foster Care Entry
1Using Exploratory Spatial Data Analysis to
Examine Neighborhood Structure and Foster Care
Entry
Bridgette Lery, M.S. Center for Social Services
Research, UC Berkeley
This research was funded by the California
Department of Social Services, the Annie E. Casey
Foundation and the Stuart Foundation.
2Overview
- Neighborhood Research
- GIS vs. Spatial Statistics
- Techniques for Exploring Spatial Distributions
- Implications for Social Research and Practice
3Alameda County 2000-2003 First Entries
4What is spatial data analysis?
- - Examining some process in space and its
relationship to other spatial phenomena - - Used to describe relationships between points,
lines, or areas - - GIS is not spatial analysis
5Why use spatial analyses?
- Social problems occur in some place, but place
often not explicitly explored - Adjacent spatial units may be correlated (e.g.,
neighborhood rates of poverty) - Neighborhood boundaries are permeable
6Spatial Scale
7Spatial Scale
8- GeoDa
- http//sal.agecon.uiuc.edu/csiss/geoda.html
9Spatial Weights
- 0 320 trt00 ID
- 1 6
- 141 140 129 118 43 44
- 2 4
- 141 43 4 3
- 3 7
- 12 11 6 4 2 42 43
- .
- .
- .
- 314 4
- 319 318 313 312
10Connectivity
11Spatial Weights Connection Matrix
12Spatial Autocorrelation
- Correlations between spatial units
- Violates the assumption of unit independence
- Biases the statistical tests of the coefficients
13Measuring Spatial Autocorrelation
- Moran Coefficient ( 1/(N-1))
- Approximately bounded by (-1) and (1)
- Can be adapted to examine residuals from
regression models
14Positive Spatial Autocorrelation
- Adjacent units have same or similar value on
measures - If present, results in Type I errors
15Negative Spatial Autocorrelation
- Adjacent units have opposite values on measures
- If present, results in Type II errors
16Movement of People in Places
- Spatial Lags measure how characteristics of
nearby places or populations influence a local
area
17Moran Scatter Plot
18An Example
Is social organization related to neighborhood
rates of foster care entry?
19Methods
- - Cross-sectional ecological study (modeled after
Coulton et al., 1995) - - of first entries to foster care per thousand
children (Data Source California Childrens
Services Archive, CWS/CMS, 2004, Quarter 1
Extract) - Geocoded to 46 zip codes, 320 census tracts and
983 block groups in Alameda County in California
by the removal address of the child
20Independent Measures
- 3. Hispanic Immigrants
- Hispanic
- elderly
- Ratio children to adults
- 4. Asian Immigrants
- Asian
- foreign-born
- Population
- Population Density
- 1. Impoverishment
- female-headed households
- poverty
- unemployed
- vacant housing
- African American
- 2. Residential Instability
- moved past 5 years
- moved past year
- moved past 10 years
- Ratio men to women
21(No Transcript)
22LISA map
23Comparison of OLS and GLS Models
p lt .05, p lt .01, p lt .001
24Conclusions
- Overall
- ESDA can help in the examination of social
phenomena that occur in a geographic context. - In the current study
- Spatial autocorrelation is positive and
significant. Therefore, the spatial model is
better than the OLS model. - The LISA map suggests specific areas where high
foster care entry rates are clustered.
25Implications for Social Work Practice
- Research Spatial dependence can bias results in
any ecological study. - - Policy and practice Neighborhoods share and
owe some characteristics to nearby areas,
suggesting a role for policies targeted to
neighborhoods and communities.