Addressing Problems with Traditional Crime Linking Methods Using Receiver Operating Characteristic Analysis

• NCJ Number: 229349
• Journal: Legal and Criminological Psychology Volume: 14 Issue: 2 Dated: September 2009 Pages: 293-310
• Author(s): Craig Bennell; Natalie J. Jones; Tamara Melnyk
• Date Published: September 2009
• Length: 18 pages
• Publication Format: Article
• Publication Type: Research (Theoretical)
• Language: English
• Country: United Kingdom

Annotation

Based on an examination of serial rape data, this article presents arguments in support of the use of receiver-operating-characteristic (ROC) analysis as the best method for determining when crimes are linked in a serial pattern.

Abstract

The arguments focus on the fact that traditional methods of linking crime in determining a serial pattern do not take into account how linking accuracy will vary as a function to the threshold used for determining when two crimes are sufficiently similar to be considered linked. The authors further argue that the solution is to use a method of analysis that can quantify the degree of linking accuracy achieved under any given set of conditions, unbiased by threshold placement. Also, the method of analysis would ideally guide decisions about threshold placement, so as to optimize performance on the linking task. The authors believe that ROC analysis meets these criteria. ROC analysis was originally developed in the field of signal detection, but it is now commonly used to evaluate and improve decisionmaking performance in a variety of diagnostic fields. The authors have previously argued that linkage analysis can be viewed as a signal-detection problem, at least when the linking task involves the consideration of whether a pair of crimes has been committed by the same offender. The goal in the linking task is similar to the goal for any diagnostic task, which is to identify a relatively rare signal (a linked crime) against a background of noise (unlinked crimes). In addition, linking decisions of this type must often be based on ambiguous evidence, such as a high across-crime similarity score that can arise from an examination of both linked and unlinked crimes. Considered for analysis in developing the argument for ROC analysis were 27 crime-scene behaviors exhibited in 126 rapes, which were committed by 42 perpetrators. 2 tables, 2 figures, 35 references, and appended listing of the 27 variables