Creating and validating risk models [30]

Go to: Summary | Previous | Next   
Bullet points include: Forecast based on one sample (possibly sector or time specific) compared to data from another sample (possibly sector or time specific) E.g. in-sample and out-of-sample back testing But bear in mind that out-of-sample back testing does not eliminate look-back bias as we only have one past we can observe, however many times it is resampled Usually there is rolling calculation of forecasts, once model is up and running Hence ability to calculate fraction of correct predictions, and to apply statistical tests to test reliability (see later)

Contents | Prev | Next | ERM Lecture Series

Desktop view | Switch to Mobile