The CrimeStat Program:

1
The CrimeStat Program The Use of Spatial
Statistics for GIS Analysis of Transportation
and Law Enforcement Problems
Ned Levine, PhD Ned Levine Associates Houston,
TX
College Station, TX October 29, 2008
2
CrimeStat is developed under funding from the
National Institute of Justice Mapping and
Analysis for Public Safety (MAPS)
Program http//www.ojp.usdoj.gov/nij/maps The
latest version is 3.1 (CrimeStat III)
3
CrimeStat III

• Windows Program
• Statistical tool box for analysis of
  events/incidents
• Can input dbf, shp, dat, and Ascii files
• Can output to ArcGIS, ArcView, MapInfo,
  AtlasGIS
• Surfer for Windows, Spatial Analyst, Maptitude,
• and many other desktop GIS packages
• It is NOT a Geographic Information System. It
• works with a GIS to analyze spatial patterns.

4
(No Transcript)
5
Data Input

• Requires primary file
• Points with X and Y coordinates
• Can be actual points (e.g., incident locations)
• Can be pseudo-points (e.g., census tract
  centroids)
• Allows secondary file
• Same coordinate system as primary file
• Permits weighting of points
• Intensities (e.g., number of incidents at
  location)
• Weights (e.g., particular sub-population)
• Can use reference file
• Either input regular or irregular grid
• or calculates grid from corners

6
Coordinates and Distance

• Spherical
• Spherical Great Circle arcs
• Projected
• Two-dimensional Euclidean distance
• Network
• Along road network

7

• Spatial Statistics Routines
• Spatial Description
• Spatial distribution
• Distance analysis
• Hot spot analysis
• Spatial Modeling
• Interpolation
• Journey-to-crime analysis
• Space-time analysis
• Crime travel demand modeling
• Significance Testing
• Monte Carlo Simulation
• Outputs Graphical Objects

8
Today, Ill talk about

• Spatial Description
• Hot Spot Analysis
• Crime Travel Demand Modeling

In relation to transportation law enforcement
analysis
9
Description of the Spatial Distribution Identifyin
g overall patterns in the data
10
Example Analyzing Crashes in Metropolitan
Houston
11
(No Transcript)
12
(No Transcript)
13
High-risk Behaviors 1999-2001

• Region U.S.
• Speeding 39 13
• Failing to yield 20 19
• Failing to stop 11 9
• Running a red light 8 5
• DUI/DWI 7 7
• (However, involved in 37 of fatal crashes)
• Following too close 3 3
• Improper turn 2 2
• National Safety Council. U.S. average for
  1999. Fatal and injury crashes only

14
High-risk Populations

• Region U.S.
• Teenagers 21 16
• (9 of driving age population/
• 5 nationally)
• (17 in fatal crashes /14 nationally)
• (19 in incapacitating injury crashes)
• Elderly (65) 8 8
• (9 of driving age population/
• 13 nationally)
• (11 in fatal crashes/13 nationally
• 9 in incapacitating injury crashes)
• Male drivers 79 60
• (Compared to 58 for females
• 1.7 times more likely to be in fatal crash/
• 1.4 nationally)

15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
(No Transcript)
26
Kernel Density Interpolation Depiction of
Overall Pattern
27
(No Transcript)
28
(No Transcript)
29
Uniform
Quartic
Normal
Negative Exponential
Triangular
30
(No Transcript)
31
Duel Kernel Density Interpolation Depiction of
Overall Pattern Relative to Baseline Variable
32
(No Transcript)
33
Example Analyzing Vehicle Thefts in Houston
34
(No Transcript)
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
(No Transcript)
41
(No Transcript)
42
(No Transcript)
43
Hot Spot Tools Finding Clusters of Incidents

• Mode
• Fuzzy mode
• Kernel density interpolation
• Nearest neighbor hierarchical clustering
• Risk-adjusted Nnh
• STAC
• K-means
• Anselins Local Moran

44
Mode Routine Counting incidents at a point
45
(No Transcript)
46
(No Transcript)
47
Nearest Neighbor Hierarchical Clustering
(Nnh) Hot spots identified by Nearest Neighbor
Criteria
48
(No Transcript)
49
(No Transcript)
50
(No Transcript)
51
(No Transcript)
52
(No Transcript)
53
(No Transcript)
54
Crash Risk Analysis Comparing Incidents to a
Baseline
55
(No Transcript)
56
(No Transcript)
57
(No Transcript)
58
(No Transcript)
59
(No Transcript)
60
Interpolation or Hot Spots? Overall Pattern vs.
Precision
61
(No Transcript)
62
(No Transcript)
63
(No Transcript)
64
(No Transcript)
65
(No Transcript)
66
(No Transcript)
67
(No Transcript)
68
Crime Travel Demand Modeling Linking origins with
destinations
69
Crime Travel Demand Model
Crime Inventory
Contextual variables
Data gathering
Possible scenarios
Networks
Trip generation
Trip distribution
Model
Mode split
Network Assignment
70
Scenario Analysis Modeling Drunk Driving Crashes
71
(No Transcript)
72
Purpose of Project

• Examine Driving-While-Intoxicated (DWI) trips
  that
• end in motor vehicle crashes
• Model predictors of DWI crash trips with goal of
  intervening
• to reduce the number of crashes through
• Targeting high-risk individuals and neighborhoods
• Targeting high crash locations

73
Data Sources

• 862 DWI crashes that occurred in Baltimore County
• between 1999 and 2001 in which both the crash
  location and
• the offenders address was known.
• These represent 25 of all DWI crashes that
  occurred in
• Baltimore County during the period
• Geography used were traffic analysis zones (TAZ)
• Combined with demographic, employment, and land
  use
• data as predictor variables

74
(No Transcript)
75
(No Transcript)
76
(No Transcript)
77
(No Transcript)
78
(No Transcript)
79
(No Transcript)
80
DWI Crash Origin Model
Significant variables
Population () strong Percent White
() strong Rural () Number of Liquor Stores
() Number of Bars () Area of the TAZ (-)
81
DWI Crash Destination Model
Significant variables
Population () strong Number of Bars ()
strong Commercial acreage () Beltway
(freeway) passed through () Area of the TAZ (-)
82
(No Transcript)
83
(No Transcript)
84
(No Transcript)
85
Concentration of DWI Offenders and Crash Locations
In 15 TAZs, 16 of DWI crash offenders lived In
19 TAZs, 21 of DWI crashes occurred
86
(No Transcript)
87
(No Transcript)
88
Targeting High Risk (Origin) Zones

• Intervention involves three policies for the 15
  high risk zones
• Dont Drive while Drinking campaign
• Increase use of ignition interlock devices
• Conduct interventions with drivers convicted of
  DWI driving

89
Simulating Successful Intervention in High Risk
Zones
Scenario Reduce DWI crashes originating from
these zones by 20
Result The total number of predicted DWI
crashes are reduced by 3.5
90
Targeting High Crash (Destination) Zones
Intervention involves three policies for the 19
hot spot zones 1. Concentrated
enforcement 2. Providing of para-transit
services for drinkers 3. Provide engineering
fixes to roadway elements
91
Simulating Successful Intervention in High Crash
Zones
Scenario Reduce DWI crashes occurring in these
zones by 20
Result When combined with reduction in offender
origins, total number of DWI
crashes decreased by 7.5
92
Comparing DWI Crash Trips Before After
Intervention
Annual Number of Expected Crashes
Inter-zonal
Intra-zonal
Without intervention
226
59
With intervention
207
57
Expected Change
-19
-2
93
(No Transcript)
94
(No Transcript)
95
CrimeStat is available at
http//www.ipcsr.umich.edu/crimestat