Creating and validating risk models [41]

Go to: Summary | Previous | Next   
Bullet points include: Back-testers attempt to mitigate this problem by using out-of-sample testing Construct the model using only data available by the point in time when the model is deemed to be applied, i.e. use a different parameterisation of the model for different time periods, the model evolving through time as new information is exposed Reduces, but does not eliminate look-back bias In the background, there are lots of other possible risk model formulations, each with its own parameterisation, that we could have used Only the ones that have exhibited good out-of-sample back test properties are likely to see the light of day (whether selection is conscious or unconscious)

Contents | Prev | Next | ERM Lecture Series

Desktop view | Switch to Mobile