Various parametric models used for prediction are described, followed by a consideration of the conditional likelihood aproach in Cox's 'semi-parametric' model, along with some of the difficulties. Alternate models subsequently developed are also examined, and some general problems in analysis of covariance and how they relate to the treatment of censored life data are explored. The data used to illustrate the application of the models was taken from an extensive followup study of the postrelease behavior of 108 parolees in Connecticut for 15 months after release. The effects of covariates such as age, income, and drug use on length of time to rearrest were examined. The records of two correctional institutions were compared after adjusting for nonhomogeneity of covariate values. Tabular data and 66 references are provided. (Author abstract modified)