Model Selection
From Open Risk Manual
Definition
Model Selection is a stage of the Quantitative Model Development process that aims to select the optimal model from a range of competing examples that have been shown to pass all relevant diagnostic tests.
The selection process might (depending on the nature of the model) employ various tools, such as Diagnostic Tests, automated procedures such as multiple regression, stepwise regression etc.
Issues and Challenges
- Among statistical models, alternative models may belong to altogether different model classes, most commonly they concern different variables (regressors) or make different transformations of such variables.
- Among pricing models, alternative models may make different assumptions about the underlying market process thereby leading to different modeled prices
