© 1993 by Institute of Mathematics and its Applications
Reject inference applied to logistic regression for credit scoring
Department of Statistics, University of Leeds
Using logistic regression to model probabilities of group membership applied to credit scoring, appropriately adjusted posterior probabilities are used to reflect prior probabilities of assignment to each group and differential costs of misclassification. A reject-inference procedure based on iterative reclassification is adapted to this framework, to produce a modified set of parameter estimates reflecting the fractional allocation of the rejects.