ABSTRACT
This study examined age differences in (1) the latent structure of the SAVRY, (2) the correspondence among structured professional judgments of risk and an empirically-derived measure, and (3) the SAVRY's predictive validity for probation outcomes. The sample consisted of 177 juveniles placed on probation. Analyses were conducted on two separate age groups: ages 13–15 and 16–18. We found that the SAVRY measured “risk” similarly across age (i.e., invariant latent structures) and that structured judgments of risk corresponded with the empirically-derived measure across both age groups. However, findings regarding the predictive validity of the SAVRY for probation outcomes were mixed. These findings provide strong support for the measurement and construct validity of the SAVRY across age and highlight the need for additional research on age differences in the strength of its predictive validity.
Acknowledgment
We would like to thank Gina Vincent for her consultation during the data collection and analysis phase of this project.
Funding
Preparation of this manuscript was supported by Grant #11-98149-000-USP funded by the John D. and Catherine T. MacArthur Foundation. The authors are grateful for their support. However, the research results reported and the views expressed in the paper do not necessarily imply any policy or research endorsement by our funding agency.
Notes
1 Prior to implementation, each juvenile probation officer completed a two-day training workshop for the SAVRY that covered information about the trajectories of youth offending, research on risk factors, and the test scoring criteria. The workshops included rating the SAVRY for two case vignettes, which were reviewed and discussed as a group. All probation officers are required to hold at least a bachelor's degree in a social services–related field.
2 Fifteen cases were missing data for one or more of the SAVRY domains. Therefore, all CFA models included 162 juvenile probationers that had valid data for all four SAVRY domains. There were no significant differences in age group, race, offense type, violence SRR, nonviolent delinquency SRR, probation completion, and time on probation across case with valid domain data (n = 162) and cases without valid domain data (n = 15).
3 Root Mean Square Error of Approximation (RMSEA) values of 0.05 or less indicate a close model fit, and values between 0.05 and 0.08 indicate an adequate model fit (Browne & Cudek, Citation1993). The Comparative Fit Index (CFI) and the Tucker Lewis Index (TLI) measure the covariation among the observed items (Bentler, Citation1990). Both TLI and CFI range between 0 and 1; value greater than .90 indicate an acceptable model fit (Browne & Cudek, Citation1993). Finally, the chi-square test of model fit (χ2) indicates whether the specified model's covariance structure is significantly different from the observed covariance matrix (Byrne, Citation2001). A non-significant p-value is desirable. However, the usefulness of χ2 as a measure of model fit is questionable due to its sensitivity to sample size and the distribution of the items (Yu, 2002).
4 Four cases were missing probation outcome and were excluded from these analyses.