Stress testing incorporating causal dependency [13]

Go to: Summary | Previous | Next   
Bullet points include: Suppose our final goal is the joint distribution (to calculate VaR, TVaR etc.) Getting there directly is too hard Start with marginal (stand-alone) probabilities and (some) low order conditional probabilities Inject information about how we expect the world to work By using Bayesian nets (i.e. Bayesian networks) The simpler the structure of the Bayesian nets we use, the fewer (and simpler) are the marginal and conditional probabilities that we will need to supply Even if conditional probabilities are too difficult to identify we may still be able to derive useful bounds on parts of the probability distribution

Contents | Prev | Next | ERM Lecture Series

Desktop view | Switch to Mobile