Jaccard's Heel: Radex Models of Criminal Behaviour Are Rarely Falsifiable When Derived Using Jaccard Coefficient

• NCJ Number: 238086
• Journal: Legal and Criminological Psychology Volume: 17 Issue: 1 Dated: February 2012 Pages: 41-58
• Author(s): Paul J. Taylor; Ian J. Donald; Karen Jacques; Stacey M. Conchie
• Date Published: February 2012
• Length: 18 pages
• Publication Format: Article
• Publication Type: Report (Study/Research)
• Language: English
• Country: United Kingdom

Annotation

This study tests the validity of "radex" models of the facets of criminal behavior patterns based on data obtained from crime scenes and crime victims that are then subjected to a particular method of statistical analysis ("Jaccard" coefficient).

Abstract

The "radax" model of criminal behavior is graphically portrayed as a circle in which the center is the core motivation for the various behaviors manifested in the commission of a particular crime. The segments of the circle represent the various offender behaviors involved in the crime commission as determined from crime-scene data and data obtained from victims and/or witnesses. The model is explained with an example of how some criminologists have interpreted rape scenarios. The core motivation (center of the circle) is the offender's unsatisfied desire for intimacy, which has not been met through reciprocal interactions in which persons voluntarily give and receive the features of intimacy. The rape thus consists of behaviors that separately and in combination (segments of the circle) constitute an effort to obtain intimacy through force and manipulation (the center of the circle). In a typical study of criminal behavior, the extent to which offender behaviors co-occur is measured by using the Jaccard association index, a coefficient calculated as the number of occasions two behaviors co-occur over a number of similar crimes as a proportion of the occasions when at least one of the behaviors occurs. These co-occurrences are then examined using a multidimensional scaling method (e.g., smallest space analysis, SSA-1). In the current study, data equivalent to that examined in previous studies, as well as artificial data, that varied on four parameters were examined using the aforementioned procedure. The authors conclude that study findings overwhelmingly indicate that this approach does not provide a full test of radex models of criminal behavior. 1 table, 4 figures, and 40 references