Abstract
Prediction of success in a special program is often used to guide the selection of program participants and the treatment that they receive. Statistical analysis can be instrumental in developing such predictions, but if the analysis is based on data that resulted from the special selectivity built into the program, the predictions can be misleading and validation tests can be less informative than is commonly realized. This argument is developed and applied to problems in criminal justice administration. Data from pretrial release in federal district court is used as an illustration.

This publication has 7 references indexed in Scilit: