Difference between revisions of "Model Calibration"

From Open Risk Manual
 
(No difference)

Latest revision as of 13:19, 19 October 2020

Definition

Model Calibration refers to an essential component of the Model Development process, whereby available data or Expert Judgement are used to fit free parameters or other aspects of a given Risk Model.

  • The term Model Calibration is also used in the context of choosing a confidence level, scaling parameter or any other choices required to transform risk model outputs in specific actionable metrics.
  • The term Model Fit or Model Estimation is closely related when the underlying model is purely statistical

Examples

  • We calibrated our credit score ranges to the internal ratings scale
  • We calibrated our new interest rate model to swap and swaption data
  • We calibrated our risk limits to the 99% confidence level output of the new VaR system