Validation of the Correctional Offender Management and Profiling Alternative Sanctions (COMPAS)
This dissertation investigated the internal consistency and structural validity of the 19 Needs scales and three of the four Risk scales from the COMPAS Core Needs and Risk factor models, using a community-based, diverse sample of male and female offenders referred for diversion programming. It was predicted that Cronbach’s coefficient alphas would provide support for the internal consistency of some but not all of the COMPAS Needs scales, based on theory and past research. Results from confirmatory factor analysis (CFA) were expected to indicate poor structural validity for the Needs and Risk models. Several alternative models were tested by using CFA, including a 19 factor Needs model, an 18 factor Needs model, a two-factor Risk model (consisting of the General Recidivism and Violent Recidivism Risk scales), a one-factor Risk model, a General Recidivism Risk factor model, and a reduced one-factor model. It was predicted that the one-factor Risk model would provide an optimal fit to the data. Overall, results suggested poor structural validity for both Needs models as well as for all four Risk models. There were mixed findings for the internal consistency and structural validity of the 19 Needs scales, and particularly poor structural validity for the Violent Recidivism Risk scale. Clinical implications are discussed for correctional agencies relying on the COMPAS to facilitate decision-making about the release and risk management of defendants.
Pierson, Ashley M, "Validation of the Correctional Offender Management and Profiling Alternative Sanctions (COMPAS)" (2018). ETD Collection for Fordham University. AAI10929777.