03.03.2015 Views

2000115-Strengthening-Communities-with-Neighborhood-Data

2000115-Strengthening-Communities-with-Neighborhood-Data

2000115-Strengthening-Communities-with-Neighborhood-Data

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

378 <strong>Strengthening</strong> <strong>Communities</strong> <strong>with</strong> <strong>Neighborhood</strong> <strong>Data</strong><br />

similar values compared <strong>with</strong> its neighbors and, when summarized<br />

across all areas, is proportionate to a global indicator of spatial association.<br />

Popular LISA statistics are local Moran’s I and Gi/Gi*, which can<br />

be used to map hotspots, coldspots, and/or spatial outliers (areas <strong>with</strong><br />

values that are inversely related to those of their neighbors, as explained<br />

above) (Anselin, Sridharan, and Gholston 2007). Programs in which<br />

these maps can be created include OpenGeoDa, R (R-Geo packages, such<br />

as GeoXP), the Python-based Spatial Analysis Library PySAL (all free<br />

and open source), and Esri’s spatial statistics toolbox.<br />

As mentioned above, street segments have frequently been used as units<br />

of analysis to which points are aggregated. Several recent methods extend<br />

existing cluster techniques to networks, including kernel density mapping<br />

(Okabe, Satoh, and Sugihara 2009); local Ripley’s K function (Okabe and<br />

Yamada 2001); and LISAs (local Moran’s I and Gi/I*) (Yamada and Thill<br />

2010). These improved methods allow analysts to generate cluster maps<br />

at the street segment level that can be visualized through varying street<br />

colors, widths, or heights. Existing software to apply some or all of these<br />

methods include SANET (for ArcGIS), GeoDaNet, and PySAL 1.4. These<br />

programs also allow users to compute street network distances between<br />

origins and destinations; 9 an analyst could determine, for example, if services<br />

are <strong>with</strong>in reach of target area residents. GeoDaNet also generates<br />

spatial weights based on network distances. Another program, the Urban<br />

Network Analysis Toolbox for ArcGIS (recently released by the Massachusetts<br />

Institute of Technology’s City Form Lab), enables the analysis of<br />

point patterns using graph methods <strong>with</strong> street networks.<br />

Several recent exploratory methods add a time dimension to the traditional<br />

cross-sectional spatial cluster methods to detect local clusters across<br />

space and time. A popular local test for detecting space–time hotspots of<br />

event points is the space–time permutation scan by Kulldorff et al. (2005)<br />

implemented in the SaTScan software (Kulldorff 2010). This method<br />

detects events (such as crimes) that occurred in a similar location at a<br />

comparable time and tests the significance of these space–time hotspots<br />

by using a Monte Carlo permutation approach [for a comprehensive overview<br />

of statistical analysis of spatial points patterns, see Diggle (2003)].<br />

Rey (2001) integrated LISAs <strong>with</strong> classic Markov transition matrices<br />

to test if joint transitions of an area and its neighbors remain significant<br />

for a given set of time periods. One of the questions this method can<br />

address is if area-based hotspots (e.g., of poverty, crime, or house prices)<br />

persist over time. The LISA Markov is implemented in PySAL. 10

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!