Abstract
This study suggests that the use of traditional statistical models can increase the already difficult task of discriminating between specific criterion populations, e.g., parole successes and parole failures. It would seem valuable for correctional researchers to supplement these standard models and approaches with others that allow for a relaxation of the assumptions of linearity and homoscedasticity, in particular. This should make it possible to uncover relationships, between independent and dependent variables, that might otherwise remain obscured. The potential value of this more flexible approach is seen in the present study-one which suggests that psychological characteristics of selected offenders may account for considerably more variance, in measures of parole perform ance, than has been reported to date.
Get full access to this article
View all access options for this article.
