2 54 CHAPTER 2 that the variability among studies that could not be explained by chance was small. 2.4.3.1.1 Moderators The data provided sufficient information to run a moderator analysis only on location type, which showed no significant effect (Qbetween = 0.1, df = 1, p = .71). There was not enough information to run analyses on continuous moderators. 2.4.3.2 Predictive validity of change scores for violent (including sexual) recidivism For violent (including sexual) recidivism, the fixed-effect weighted Cohen’s d was 0.14, 95% CI [0.05, 0.24], N = 6,234 based on 6 unique samples. The Q indicated that the variability across studies was not significant (Q = 10.54, p = .06). The I2 (52.6%) indicated that the amount of variability in effect sizes among studies that could not be explained by chance was moderate. 2.4.3.2.1 Moderators The data provided sufficient information to run a moderator analysis only on location type, which showed no significant effect (Qbetween = 1.0, df = 1, p = .31). There was not enough information to run any analysis on continuous moderators. 2.4.3.3 Predictive validity change scores for any recidivism The fixed-effect weighted Cohen’s d on any recidivism was 0.10, 95% CI [-0.01, 0.22], N = 1,466, based on 4 unique samples. The Q indicated that the variability across studies was not significant (Q = 1.04, p = .79). The I2 (<0.00%) indicated that there was no variability among studies that could not be explained by chance. There was not enough information to run any moderator analysis. 2.4.4 PREDICTIVE VALIDITY OF CHANGE SCORES CONTROLLED FOR STATIC AND INITIAL DYNAMIC SCORES For this part of the analyses, only studies that reported the predictive validity of change scores, controlled for both static and initial dynamic scores, were included. The average follow-up time for the included studies was 87.2 (range = 36–147) months for studies predicting sexual recidivism, 57.4 (range = 36–112) months for studies predicting violent (including sexual) recidivism and 63.9 (range = 36–89) months for studies predicting any recidivism (sexual, violent, or any). Observed recidivism rates were 11.3% for sexual recidivism, (6 samples, n = 1.980), 10.5% for violent (including sexual) recidivism (5 samples, n = 4,168) and 29.7% for any recidivism (3 samples, n = 1,172). Table 2.4 summarizes the effect sizes of the predictive validity of change scores on dynamic risk assessment instruments corrected for both static and initial dynamic score on the three types of recidivism. Effect sizes of the individual samples are published in the Appendix B4 (available in the online supplemental materials).
RkJQdWJsaXNoZXIy MjY0ODMw