Extending a Logistic Approach to Risk Modeling through Semiparametric Mixing

摘要: The new proposal of the Basel Committee on banking regulation issued in January 2001 allows banks to use internal ratings systems classify firms. Within this context, main problem is find a model that fits data as well possible, but one also provides good prediction and explicative capabilities. In paper, our aim compare two kinds classification models applied creditworthiness using weighted error performance function: standard logistic mixed model, adopting, respectively, parametric semiparametric approach. former related assumption an i.i.d. hypothesis, it often necessary consider possible presence unobservable heterogeneity characterizes microeconomic data. To better phenomenon, we defined random effect avoiding assumptions upon distribution. This leads likelihood integral kernel density with respect mixing density, which has no analytical solution. can be obviated by approximating finite sum densities, each characterized different set parameters. discrete nature helps us detecting non-overlapping clusters homogeneous values insolvency risk, classifying firms these means estimated posterior probabilities component membership.