Semiparametric Regression Models with Applications to Scoring: A Review

Abstract

Semiparametric regression models have been proposed in the econometric literature as a trade-off between the simple but easily implementable and interpretable parametric models and the flexible but structure free smoothing techniques. Some semiparametric models for binary response with possible application to scoring data are reviewed: single-index models, generalized partially linear models, generalized partially linear single-index models, and multiple-index models. All these models are extensions of the classical logistic regression.

Keywords:

• Bandwidth choice
• Index regression
• Nonparametric regression
• Scoring
• Semiparametric models

Mathematics Subject Classification:

• Primary 62H30
• Secondary 62G08, 62J12

Acknowledgment

The author thanks Jean-Loïc Berthet for helpful suggestions on a preliminary version of this article.

Notes

¹We thank Etienne Marot for providing us with the SAS macros for drawing the CAP curves.