March 6th, 2002, 11:40 pm
If you check the thread entitled 'my VaR methdodlogy' by myself , you'll see that Aaron suggests mapping individual instruments to market factors, because those will produce more stable variance/cvariance matrices. Now for MC sims, those are obviously irrelevant, as you're not going to use them. However, if you have to estimate the parameters of the distribution you're going to use, you might be better off with the interest rates, as those might - and someone more senior needs to confirm this- produce more accurate parameters. In any cas, I's go with the interest rates and the spreads, rather than with the prices.