Transaction Search Form: please type in any of the fields below.
Date: August 17, 2024 Sat
Time: 2:16 am
Time: 2:16 am
Results for lsi-r
1 results foundAuthor: Zhang, Sheldon Title: COMPAS Validation Study: Final Report Summary: COMPAS (Correctional Offender Management and Profiling Alternative Sanctions) is a computerized database and analysis system designed to help criminal justice practitioners determine the placement, supervision, and case-management of offenders in community and secure settings. The California Department of Corrections and Rehabilitation contracted with the University of California, Los Angeles (UCLA), and San Diego State University (SDSU) to validate the instrument in terms of its ability to identify treatment needs among inmates as well as predict various recidivism outcomes. A total of 91,334 parolees who had been assessed with COMPAS prior to release were included in the study sample. Of these, roughly 60,000 had been on parole for at least 12 months and the remainder had been on parole for at least 24 months. Characteristics of the study subjects closely paralleled those of the general parolee population in California. The COMPAS needs scales were evaluated in terms of their reliability over time (test-retest coefficients) and the extent to which their constituent scales correlated with relevant counterparts on the Level of Service Inventory-Revised (LSI-R) scale (concurrent validity). To accomplish this, the COMPAS was administered twice to 75 inmates at the California Institute for Men (CIM) located in Chino, California, over a span of approximately two weeks. To establish concurrent validity, the LSI-R was also administered at the same time points. The COMPAS scales showed extremely high test-retest reliability, ranging from .70 to 1.00. The perfect and near-perfect correlations obtained for many of the scales appear to be driven by the fact that these scales were coded directly from the inmates’ Central Files. Overall, the average test-retest correlation coefficient for the COMPAS scales was .88. Of the 18 scales making up the core of the COMPAS assessment, nine appeared to measure identical or similar constructs with scales found in the LSI-R. For six of these scales (Criminal Involvement, Criminal Associates/Peers, Substance Abuse, Financial, Vocational/Educational, and Housing), significant and positive correlations were found between the COMPAS and LSI-R. The correlations were marginally significant for two of the scales, Family Criminality (COMPAS) with Family/Marital (LSI-R) and Criminal Attitudes (COMPAS) with Attitudes/Orientation (LSI-R), and not significant for one, Leisure (COMPAS) with Leisure/Recreation (LSI-R). Using official records data provided by the California Department of Corrections and Rehabilitation (CDCR), two major outcome measures were examined: (1) any subsequent arrest, and (2) a subsequent arrest for a violent offense. For the first measure, the overall re-arrest rate for the COMPAS sample was 56% for the first 12 months on parole and 70% for those who had been released for two years. For violent offenses, the re-arrest rates were approximately 13% and 21% in the 12- and 24-month periods following release, respectively. Receiver Operating Characteristic (ROC) curves were computed to assess the overall accuracy of the COMPAS risk scales for recidivism and violence. The ROC curve (measured in terms of the area under the curve or AUC) has become a primary measure of predictive accuracy in research instrumentation. A value of .70 is generally considered minimally acceptable. The recidivism and violence COMPAS risk scales were examined with regard to how well they predicted whether a parolee had been re-arrested (for any reason and for a violence offense) within two years of being released from prison. Both risk scales achieved levels of accuracy greater than chance, with the recidivism scale receiving an AUC value of .70, and the violence scale receiving an AUC value of .65. The risk prediction resulting from the COMPAS scales was comparable to our own risk prediction models using existing electronic records maintained by CDCR. We conclude that the COMPAS scales have high test-test retest reliability and moderate concordance with select LSI-R scales (with significant or marginally significant associations with eight of the nine scales that overlap with the LSI-R). With regard to the predictive validity of the recidivism and violence COMPAS risk scales, the general recidivism risk scale achieved an AUC value of .70, which is the conventional threshold for acceptability; the violence scale, however, fell short of this threshold. Details: Sacramento: California Department of Corrections and Rehabilitation, Office of Research, Adult Research Branch, 2010. 31p. Source: Internet Resource: Accessed December 15, 2010 at: http://www.cdcr.ca.gov/adult_research_branch/Research_Documents/COMPAS_Final_Report_08-11-10.pdf Year: 2010 Country: United States URL: http://www.cdcr.ca.gov/adult_research_branch/Research_Documents/COMPAS_Final_Report_08-11-10.pdf Shelf Number: 120517 Keywords: LSI-RParoleesRecidivismRisk Assessment |