Second, we tested the hypothesis that each recidivism risk factor is associated with recidivism using recidivism 3 years after release as the dependent variable and the sociodemographic factors and risk factors associated with the C8 as the independent variables. A chi-square or Fisher's exact test was conducted [30 (link)]. The difference between the population mean test was performed for age. To calculate the effect size, Cohen's d was used for age and the φ coefficient was used for other variables.
Diagnostic categories that were found to be meaningful in the Kaplan–Meier survival analyses and risk factors were examined with Cox proportional hazards models to consider their value for recidivism. Age was included in the analysis after classification with reference to descriptive statistics.
BellCurve (Social Survey Research Information Co., Ltd., Japan) for Excel (Version 4.04) was used for all statistical analyses.