Correlation, co-dependency and risk aggregation [32]

Go to: Summary | Previous | Next   
Bullet points include: Relatively simple. Generate a vector, x, of correlated Gaussian random variables. And then transform as follows, where inverse cdf of the target marginal is: Simplest way of generating correlated Gaussians if rho i,j = rho for all i <> j is (as we have seen already) to simulate a Gaussian common factor X and independent random variables epsilon 1, ..., epsilon m and then form: If rho i,j vary for i <> j then use Cholesky decomposition, see e.g. Nematrian website

Contents | Prev | Next | ERM Lecture Series

Desktop view | Switch to Mobile