1 of 36

CrimeStat IV

Susan C. Smith

Christopher W. Bruce

Revised by: Thomas Mueller

2 of 36

About CrimeStat PowerPoint

This PowerPoint was revised from the Susan C. Smith and Christopher Bruce’s work. The Original can be found at:
http://www.icpsr.umich.edu/CrimeStat/workbook.html

Their PowerPoint and Manual was developed for CrimeStat 3.0. This PowerPoints was revised for CrimeStat 4.0
The GeoTech Center wants to thank both of them for allowing us to use this tremendous resource.

3 of 36

Chapter Three��Spatial Distribution

4 of 36

In Chapter Three...

Spatial Forecasting
Mean and median centerpoints
Measures of variance
Analyzing a cluster
Limitations of spatial distributions

5 of 36

Introduction

Introducing Spatial Distribution

Forecasting

Part Art / Part Science

Probability

Of being right
Of being wrong

Forecasting is inherent in any spatial or temporal analysis

6 of 36

7 of 36

Spatial Forecasting

Two Step Process

Identify the target area for the next incident
Identify potential targets in the target area

8 of 36

Spatial Forecasting

Targets

Consider availability of targets in any given area

Banks, restaurants, convenience stores (vs.)
Pedestrians, parked cars, houses

9 of 36

Spatial Forecasting

Three types of spatial patterns in tactical crime analysis

Those that cluster

Concentrated in an area, but randomly dispersed

Those that walk

Offender moving in a predictable manner in distance & direction

Hybrids

Multiple clusters with predictable walks, or
Cluster in which the average points “walks”

10 of 36

Types of Spatial Patterns in Tactical Analysis

11 of 36

Tactical Crime Analysis

What is the minimum number of crimes needed for crime pattern?

What conditions must be met? Students should use examples in these explanations

12 of 36

Tactical Crime Analysis 2

Please explain the difference between a crime trend and a crime pattern?

Please explain the difference between a crime pattern and a problem?

What is the role of statistics in defining a crime pattern?

13 of 36

Tactical Crime Analysis 3

Students should attempt to find examples of crime patterns types via a web search and discuss them

Series
Spree
Hot Prey
Hot Product
Hot Spot
Hot Place
Hot Setting

14 of 36

Now Return to CrimeStat

15 of 36

Spatial Distribution

How are the crimes distributed?

Average location?
Greatest volume / concentration?
Boundaries?

Questions can be answered by looking at (points):

Mean Center - Geometric Mean
Harmonic Mean - Median Center
Center of Minimum Distance

16 of 36

Spatial Distribution

Questions can be answered by looking at (areas):

Standard Deviation of X & Y Coordinates
Standard Distance Deviation
Standard Deviation Ellipse
Two Standard Deviation Ellipse

17 of 36

Measures of Spatial Distribution

Mean Center

Intersection of the mean of the X coordinates and the mean of the Y coordinates

Mean Center of Minimum Distance

The points at which the sum of the distance to all the other points is the smallest

Median Center

Intersection between the median of the X coordinates and the median of the Y coordinates

Great if you have outliers!

18 of 36

Measures of Spatial Distribution

Geometric Mean & Harmonic Mean

Alternate measures of the mean center
Just rely on the mean

19 of 36

Measures of Concentration

Standard Deviation of the X and Y coordinates

A rectangle encloses the area in which four lines intersect: one s/d above the mean of the X axis, one s/d below the mean on the X axis, one s/d above the mean on the Y axis and one s/d below the mean on the Y axis

Standard Distance Deviation

Calculates the linear distance from each point to the mean center point, then draws a circle around one s/d from the center point.

20 of 36

Measures of Concentration

Standard Deviational Ellipse

Similar to the standard distance deviation but accounts for skewed distributions, minimizing any “extra space” that might appear in a circle

Convex Hull Polygon

Encloses the outer reaches of the series.
No points fall outside of the polygon

Outliers may greatly increase the size of the polygon

21 of 36

Analyzing a Cluster

Open CrimeStat
Click on Spatial Description tab in CrimeStat
Complete the Steps from Chapter 2 to set up the burglaryseries analysis

22 of 36

Spatial Description:�Primary File

23 of 36

Spatial Description:�Reference Grid

24 of 36

Spatial Description:�Measurement Parameters

25 of 36

Analyzing a Cluster

Select all of the checkboxes in Spatial Description from ―Mean center and standard distance, Stand Dev Ellipse and Median Center.

26 of 36

Analyzing a Cluster

For each of the checked boxes, click on the ―Save Result To on the right, choose ―ArcView ‗SHP, click ―Browse and set the directory and add a 1 in front of burglaryseries (burglaryseries1)

27 of 36

Analyzing a Cluster

Click Compute
Eight (8) shapefiles will be created

28 of 36

Eight (8) ArcGIS shapefiles will be created

2SDE: Two standard deviation ellipse
GM: Geometric Mean
HM: Harmonic Mean
MC: Mean Center
MdnCntr: Median Center
SDD: Standard Distance Deviation
SDE: One standard deviation ellipse
XYD: Standard deviation of the X and Y coordinates]

Now we need to set the projections of these new shapefiles

29 of 36

ArcGIS Pro: �Setting Projection

Click Analysis - Tools

30 of 36

ArcGIS Pro: �Setting Projection

In Geoprocessing Pane – search Define Project (the new Shapefiles do not have a projection or coordinate system yet

31 of 36

ArcGIS Pro: �Setting Projection

In Input - Browse to the burglaryseries and click it
In the Coordinate System – choose Burglaryseries (It will change it to the Lincoln – which is the Shapefiles projection

32 of 36

ArcGIS Pro: �Setting Projection

It will now have the same projection system as the original data
Click Run
Repeat the steps for the rest of the shapefiles

33 of 36

Analyzing a Cluster

Open each, format and compare

34 of 36

Discussion (page 33)

1. If the agency decides to place an unmarked patrol car in the area to respond quickly to any alarms or reports of burglaries, where would you suggest that they station it?
2. In what area would you predict the next offense is likely to occur?
3. In what area would you predict the next offense will almost certainly occur?
4. If the agency wanted to suppress the offender by saturating the area with patrol offi-cers, in what area would you recommend they concentrate?
5. If the agency wanted to station ―scarecrow cars‖ in the area to deter the offender, where would you recommend that they station them?
6. If the agency wanted to alert residents about the series, encouraging potential future victims to lock doors and hide valuables, in what area should they call or leave notices?

Answers: Bruce and Smith p.34

“These are our answers. If yours differ slightly based on your analytical judgment, that‘s fine; there are no absolute ―rights‖ and ―wrongs‖ here.

An unmarked car stationed to respond quickly to incidents would probably best be sta-tioned at one of the mean center calculations (in this case, it doesn‘t matter much, but the mean center of minimum distance would minimize response times). That would be about halfway along West Harvest Drive.

2. Assuming we are correct that the series doesn‘t ―walk,‖ the next offense is most likely to occur within a single standard deviation of the mean center. The standard deviation ―rectangle,‖ the single standard deviation ellipse, and the standard distance deviation all provide good estimates of the densest concentration.

3. The next offense will almost certainly occur within the two standard deviation ellipse or the convex hull polygon, unless the offender does something completely new for his next burglary. These of course have the disadvantage of being larger and thus harder to con-centrate resources within.

4. Saturation patrol would best be concentrated in the single standard deviation ellipse, because it encloses a majority of the incidents and conforms best to the street geography in this part of the city.

5. This is a bit of a trick question because it‘s not fully answerable with these CrimeStat cal-culations. But note that the standard distance deviation encompasses a neighborhood with few entries and exits. An offender would almost certainly have to pass cruisers sta-tioned at four locations: the intersection of NW 12th St and W Highland Blvd; NW 13th St and W Fletcher Ave; NW 1st St and W Fletcher Ave; and NW 1st St and W Highland Blvd.

6. To reach all potential targets, the two standard deviation ellipse would be the best choice.”

35 of 36

Cautions & Caveats

You generally can’t do this by hand

Wouldn’t account for multiple incidents at a single location
Larger series or large volumes of crime would be nearly impossible to interpret on your own
CrimeStat can be precise; you cannot (usually)

Nothing should replace your experience, intuition and the obvious (see Figure 3-7)

36 of 36

Figure 3-7: An unhelpful spatial distribution. The mean center, standard deviation ellipse, and standard distance deviation circle are technically correct, but they miss the point of the pattern, which is that it appears in two clusters. The analyst in this case would probably want to create a separate dataset for each cluster and calculate the spatial distribution on them separately.