การเปรียบเทียบการจัดอันดับและมูลค่าความเสี่ยงระหว่างตัวแบบ โพรบิทแบบภาวะภัยและตัวแบบเกาซ์เซียนคอพพูลาโพรบิทแบบภาวะภัย / ศรัณยา สมทรง = A Comparison on Ranking and Value at Risk between the Hazard Probit Model and the Hazard Probit with Gaussian Copula Model
The objective of this research is to compare the ranking of the fitted scores and the value at risk (VaR) obtained from two credit scoring models: hazard probit model and hazard probit with Gaussian copula model. The experiment is done under the following conditions. The data set contains data in multiple periods. The dependence variable is a binary variable. The observations of the dependence variable in the same period are correlated by a Gaussian copula factor at = 0.1, =0.3, =0.5 and = 0.7. The observations of the dependent variables from different periods are independent. There are two independence variables, which are simulated from the standard normal distribution. The numbers of the periods are 3, 6, 9, and 12. The number of data is 1,000 per periods. The performance measure for the credit score ranking is the rank correlation obtained from the scoring model and the data model. On the other hand, the performance measure for the credit value at risk (CVaR) is the difference of the CVaR obtained from the risk estimation model and that from the data model. The comparison is performed by testing hypotheses on the difference of the performance measures between the two models at significant levels of 0.05. From the analysis, the fitted credit scores from the two models are not significantly different in most cases, while the values at risk from the two models are significantly different in all cases.
