Random Sampling of Correlated Parameters - a Consistent Solution for Unfavourable Conditions

Two methods for random sampling according to a multivariate lognormal distribution �?? the correlated sampling method and the method of transformation of correlation coefficients �?? are briefly presented. The methods are mathematically exact and enable consistent sampling of correlated inherently positive parameters with given information on the first two distribution moments. Furthermore, a weighted sampling method to accelerate the convergence of parameters with extremely large relative uncertainties is described. However, the method is efficient only for a limited number of correlated parameters.