The Structure of Violent Behavior

Abstract

Common difficulties in measures of violent behavior are their multidimensional nature, the nonempirical ordering of violent acts, the inclusion of undiscriminating items, and differential precision of measurement across the range of seriousness. Data collected on a sample of 250 male Scottish prisoners, using the MacArthur Community Violence Screening Instrument, are examined using confirmatory factor analysis and item response theory (IRT) analysis. A hierarchical model with a coherent superordinate factor overarching two lower order factors—violence based on weapon use and violence without weapon use—provide the best fit to the data. The ordering of the items on the scale does not conform to the order indicated by IRT analysis. Items vary considerably in the amount of information they provide, with the precision of measurement of the total scale being poor at high levels of the scale. Ways of improving the measurement of violence are discussed.