In present days, retail banks select their clients on the basis of a classification criterion aiming at identifying people that are 'good' or 'bad' risks. The statistical problem is, therefore, how to update the rule when the observed rate of bad clients differs substantially from the one expected on the basis of the classifier. Methods to tackle the problem are known as 'reject inference'. They involve the quantification of the distortion induced by the selection mechanism on the basis of the available information. In this paper we detail the assumptions under which the distortion can be estimated or bounds for the parameters of interest can be provided.
On reject inference for binary response models
STANGHELLINI, Elena
2008
Abstract
In present days, retail banks select their clients on the basis of a classification criterion aiming at identifying people that are 'good' or 'bad' risks. The statistical problem is, therefore, how to update the rule when the observed rate of bad clients differs substantially from the one expected on the basis of the classifier. Methods to tackle the problem are known as 'reject inference'. They involve the quantification of the distortion induced by the selection mechanism on the basis of the available information. In this paper we detail the assumptions under which the distortion can be estimated or bounds for the parameters of interest can be provided.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
