|
Abstract:
|
The discriminatory power from a credit scoring model refers to its ability to differentiate between good and bad credits. It can be measured by various statistics such as Kolmogorov-Smirnov test, Information Value, Gini Coefficient from a CAP curve, Area under Curve from a ROC curve, and Conditional Information Entropy Ratio. However, the mathematical relationships among those statistics are not available in the literature. This paper demonstrates the empirical relationships among the statistics using Monte Carlo technique. The results can be used as a reference to derive the estimates for other statistics after one test (which may be a favorable test for a practitioner) is calculated during a model validation process.
|
