The Measurement of “Deviance” Under the Scrutiny of Item Response Theory

• In Event: Advances in Measurement

Abstract

Delinquency or deviance, as defined by Hirschi, has been extensively used in delinquency research and beyond. Among “acts, the detection of which is thought to result in punishment of the person committing them by agents of the larger society” (Hirschi, 1969, p.47) have been commonly considered and measured damaging other persons’ belongings, stealing, attacking, carrying or using a weapon, selling drugs, as main behavioral constituents of delinquency construct. From a primarily measurement perspective, this study aims to investigate psychometric properties of eight-item delinquency scale (Vazsonyi et al. 2008) within an item response theory framework. Based on two nationally representative samples of youth from the Republic of Georgia and Switzerland, ages 13 to 19 years (N=8,840 and 9,986, respectively), different statistical analyses were conducted, including the confirmatory factor analysis, to examine the unidimensionality assumption, while invariance assessments by sex and country as well as differential item functioning, were completed using IRT. Findings show that a unidimensional model fit the data well; next, different IRT models (2PL, 3PL) were tested, more specifically IRT-based linking, identified as one of the advanced methods for testing the tenability of cross-national assessments (DeMars, 2015, Engelhard & Perkins, 2011), to establish metric and scalar equivalence of deviance questions.

Authors

• Magda Javakhishvili, University of Kentucky

• Alexander T. Vazsonyi, University of Kentucky

• Helen Phagava, Tbilisi State Medical University & Institute of Medical Biotechnology, Georgia

• Karaman Phagava, Tbilisi State Medical University & Institute of Medical Biotechnology, Georgia

• Pierre-Andre Michaud, University of Lausanne, Switzerland