Evaluation of Recidivism Data

Abstract

Some recent papers have demonstrated the usefulness of failure rate methods in the evaluation of data from recidivism studies and other programs involving behavioral change. We extend the discussion to show the applicability of recently developed failure rate regression techniques Such methods allowing for covariates are important for construct validity, adjusting for bias in observational studies, and in identifying factors associated with successful outcomes One technique presented here allows for time varying covariates and does not require any particular parametric assumption about the time-to-failure distribution. Recidivism data from a Connecticut parolee study are used as an example