Tests for point level data
Suitable global tests to perform on geocoded point level data (i.e. property/location
precise or postcode precise data) are,
When performing these statistical tests it is more important to understand their
application and interpretation, rather than the mathematics and in-depth knowledge
of the equations they use.
Dispersion
Measures of standard deviation distance help to explain the level of dispersion
in crime and disorder data. These statistics are best used as relative measures,
comparing crime or disorder types against each other or the same types for different
periods of time. The greater the standard deviation distance, the more dispersed
the crime or disorder incidents.
|
Test: Standard distance deviation Standard distance deviation
Application: Used to measure relative levels of dispersion between crime
or the same crime types for different periods of time Used to measure relative levels of dispersion between crime
or the same crime types for different periods of time.
Interpretation of result: The greater the standard deviation distance,
the more dispersed the crime or disorder incidents. The greater the standard deviation distance,
the more dispersed the crime or disorder incidents.
|
Clustering
Testing for clustering performs the first step in revealing whether there are hotspots
of crime or disorder in incident data.
The nearest neighbour index (NNI) is a simple and quick method to apply to test
for evidence of clustering. The NNI test compares the actual distribution of the
crime or disorder data against a data set of the same sample size but where the distribution
is completely random.
If the result generated from the NNI test is 1 then the crime data is randomly
distributed. If the NNI result is less than 1 then the crime data shows evidence
of clustering. A NNI result that is greater than 1 reveals evidence of a uniform
pattern in the crime data.
To help place confidence in the nearest neighbour index result a test statistic
can be applied. This z-score test for statistical significance describes how different
the actual average nearest neighbour distance is to the average random nearest neighbour
distance. The significance of the z-score can be found in any table of standard normal
deviations. The general rule to follow is that the more negative the z-score the
greater the confidence that can be placed in the NNI result, bearing in mind that
for smaller sample sizes, z-score will be less than that for larger samples of crime
points.
|
Test: Nearest Neighbour Index (NNI) Nearest Neighbour Index
(NNI)
Application: Used to reveal whether there is evidence of clustering,
and therefore hotspots, in point data. Used to reveal whether there is evidence of clustering,
and therefore hotspots, in point data.
Interpretation of result: If the result generated from the NNI test is
1 then the incident data is randomly distributed. If the NNI result is less than
1 then the incident data shows evidence of clustering. A NNI result that is greater
than 1 reveals evidence of a uniform pattern in incident data. If the result generated from the NNI test is
1 then the incident data is randomly distributed. If the NNI result is less than
1 then the incident data shows evidence of clustering. A NNI result that is greater
than 1 reveals evidence of a uniform pattern in incident data.
Confidence measure: Test statistic (Z-score) - the more negative the z-score
the more confidence that can be placed in the NNI result.
Click here for an example which looks at applying
tests for clustering and dispersion against robbery, residential burglary and
vehicle crime data from the London Borough of Croydon.
|
Click here to return to Statistical Tests
for Hotspots
|
