Hi,I am trying to employ Importance sampling for an implementation of CreditMetrics. All examples I find are based on shifting the sampling mean of ONE factor, and then adjusting the result accordingly. My problem is that I have 10+ factors that I would like to shift. These factors are industry indices, and I assume that shifting only one of them would generate a "biased" loss distribution, since some positions (loans) are, of course, more dependant on certains indices than others.Can someone help me on this matter? How to do IS with a multivariate normal distribution?Thanks

Try this paper by Glasserman, Heidelberger and Shahabuddin.