Latest news
The JDI Latin America and Caribbean Unit - a new unit to support research on crime and citizen security, and professional development
Mailing List
Short Courses


Advanced Hotspot Analysis

Predictive Crime Mapping

Hypothesis Testing Analysis

Crime Analysis

Understanding Hotspots

Strategic Intelligence Assessments

Dates to be confirmed

Geographic Profiling Analysis

26th June - 7th July 2017

Department of Security and Crime Science

ICIAC 2012 Stream 5: Classes and Workshops

Abstracts and slides

CLASS 5C: How near is ‘near’? Change-point regression as a way to assess the spread of violence around bars

Jerry Ratcliffe, Temple University, USA

Slides: Ratcliffe

Crime scientists have long known that crime clusters near certain places such as drinking establishments, although the spatial parameters of that clustering are less well established. In other words, if violence occurs at and near bars, what counts as ‘near’?

This presentation demonstrates a methodology analysts can use to estimate a distance beyond which there is significantly less evidence of a correlation between criminogenic locations and concentrations of crime (crime spillover). In other words, it allows an analyst to calculate a more spatially quantifiable definition of ‘near’ that is relevant for their area and crime problem.

GIS techniques are used to create a series of buffers to determine the density of crime around sites. A change-point Poisson regression of the buffer midpoints is used to estimate the distance beyond which crime densities do not appear to decline significantly with increasing distance. All of this can be done with a GIS and a statistical routine available either from a dedicated free downloadable program, or available within the free software tool ‘R’.

A case study of violent crime around 1,282 bars in Philadelphia, PA, USA shows that violence is highly clustered within just 85 feet then dissipates rapidly, a pattern that is not replicated using control sites (fire stations). This is an estimate of the spatial extent of violence around bars in Philadelphia; however the technique could be used to estimate the extent of other crimes around a variety of crime-generating locations. 

CLASS 5D: Self-selection as an investigative tool - shoplifting and the non burgling burglar

Gordon Stovin, Principal Crime Intelligence Analyst, West Mercia Police

Slides: Yet to be supplied by presenter

Self selection* amongst criminals is a well documented and yet under exploited phenomenon within law enforcement. Possibly the earliest and certainly the most startling finding is that of Chenery, Henshaw & Pease who examined illegal parking behaviour in disabled bays and found that 22% of vehicles' keepers were of immediate interest to the police** (Chenery et al, 1999).

This workshop aims to build on this finding and examine some of the more recent and equally compelling literature on the subject of self selection. Here we will place an emphasis on links between types of acquisitive crime where the detection rates vary enormously.

With this knowledge we will look at the theory-led problem solving opportunities afforded by this literature to the crime intelligence analyst. Specifically when interpreting spates in acquisitive crimes such as domestic burglary or vehicle crime. We will look at using the principles of self selection alongside hypothesis testing to add greater value to traditional crime hotspots, identify other geographic areas vulnerable to future offending and develop more comprehensive suspect shortlists for undetected acquisitive offences.

* or as Roach (2007) puts it "those who do big bad things also usually do little bad things".

** compared to two percent of legally parked vehicles in disabled bays. 

CLASS 5E: Time-series analysis

Lisa Tompson, Department of Security and Crime Science, University College London

Slides: Tompson

Measuring change in any phenomena over time requires the use of distinct analytical techniques due to the intrinsic temporal ordering of the data. Time-series analysis is used when observations are repeatedly made over a number of time periods (usually 50 or more) to determine characteristics of the data and generate meaningful statistics. All time-series data have three basic parts; a trend component, a seasonal component and a random component. General trend analysis of these data is therefore complicated, and requires the application of advanced statistical techniques.

This class will introduce time-series analysis to delegates and familiarise them with the analytic process. It will start by outlining some of the fundamental principles of time-ordered data; then, the class will work through a number of real-world crime examples that involve techniques such as seasonal decomposition and ARIMA (autoregressive, integrated, moving average) modelling. A thread running through the whole class will be the types of questions an analyst might want to answer with time-series analysis, and which techniques are the most appropriate for doing so.

Delegates will get the most benefit out of this class if they have some foundation knowledge of statistical testing. An awareness of programming would be helpful, but not essential.

Page last modified on 30 jan 13 17:08