Tests for area aggregated data
If crime and disorder data is only available as aggregate counts (representing
the number of events within a certain geographic area, e.g. enumeration districts
or beats) an appropriate group of methods to apply to test for evidence of hotspots
are spatial autocorrelation tests.
Spatial autocorrelation techniques test for the condition of whether the distribution
of events are related to each other. Where events are clustered or where events close
together have similar values than those that are further apart, positive spatial autocorrelation
is said to exist.
Suitable tests to perform on area aggregated data are,
The Moran’s I and Geary’s C tests can be performed using CrimeStat
Global indication of the evidence of hotspots (Moran’s I).
Moran’s I statistic works by comparing the value at any one location with the value
at all other locations. Moran’s I requires an intensity value for a crime or disorder
point, often represented as the centroid of the geographic boundary area. This point
is then assigned an intensity value. For crime and disorder applications this intensity
value will be the count of crimes within that geographic area. The Moran’s I result
varies between –1.0 and +1.0. Where points that are close together have similar values,
the Moran’s I result is high.
|
Test: Moran’s I
Application: Used to reveal whether there is evidence of clustering,
and therefore hotspots, in geographic boundary aggregated data. Best applied to give
a global indication of the evidence of hotspots. Used to reveal whether there is evidence of clustering,
and therefore hotspots, in geographic boundary aggregated data. Best applied to give
a global indication of the evidence of hotspots.
Interpretation of result: The Moran’s I result varies between –1.0 and
+1.0. Areas close together that have similar values will have a high Moran’s I result.
|
Test to describe differences at the local level (Geary’s C).
Geary’s C statistic is best applied to describe differences at the local level.
Geary’s C statistic is a measure of the deviations in intensity values of each point
with one another. The values of C typically vary between 0 and 2, where values less
than 1 indicate evidence of positive spatial autocorrelation and values greater than
1 indicate evidence of negative spatial autocorrelation.
|
Test: Geary’s C
Application: Used to reveal whether there is evidence of clustering,
and therefore hotspots, in geographic boundary aggregated data. Applied to describe
differences at the more local level. Used to reveal whether there is evidence of clustering,
and therefore hotspots, in geographic boundary aggregated data. Applied to describe
differences at the more local level.
Interpretation of result: Values of C vary between 0 and 2. Values of
less than 1 indicate that incident data are clustered or where events close together
have similar values than those that are further apart. Values greater than 1 indicate
that events close together are dissimilar. Values of C vary between 0 and 2. Values of
less than 1 indicate that incident data are clustered or where events close together
have similar values than those that are further apart. Values greater than 1 indicate
that events close together are dissimilar.
Click here for an example which
looks at applying spatial autocorrelation tests for clustering against all
crime, robbery, residential burglary and vehicle crime counts at the enumeration
district level for the London Borough of Croydon.
|
Click here to return to Statistical Tests
for Hotspots
|
