Evaluations involving nonrandom assignment to treatment or control groups are vulnerable to an accidental or intentional confounding of a selection effect with the treatment effect. The resulting selection bias is compensated with two techniques, discriminant analysis and base expectancy...