Considering Measurement Invariance in Self-reported Offending: Replication of Prior Research Using Modified Self-Reported Offending Scores Adjusted for Differential Item Functioning

• In Event: School Safety, Juvenile Justice, and Research Methodologies: Navigating Policy and Practice

Abstract

Self-report measures have been heavily relied on to capture one of the most important outcomes of criminological research – offending behavior. However, the self-report method has been criticized for various reasons (e.g., honesty, recall, social desirability, etc.). A particularly salient concern is that self-report items may not measure offending the same across groups, that is, it is not invariant across theoretically and methodologically relevant groups of individuals (e.g., race/ethnicity and biological sex). However, recent research (e.g. Ray, Baker, & Peck, 2024) has shown that invariance in self-reported offending can be accounted for using factor scores derived from moderated non-linear factor analysis (MNLFA) that adjust for differential item functioning (DIF) on observed factors, including both categorical and continuous measures of those factors. In the current study, we replicate prior research findings that measure self-reported offending using additive scores, mean scores, or variety scores using DIF-adjusted scores. We compare results using DIF-adjusted scores to those using non-DIF-adjusted scores. Preliminary results suggest that scholars may have come to different conclusions had they used DIF-adjusted scores.

Authors

• James Ray, University of Central Florida

• Thomas Baker, University of Central Florida