Why may an analysis that makes even a rough estimate of the probability distribution of project NPV or IRR be infinitely more valuable to management than a point estimate Why are scenario and especially decision tree analysis particularly valuable with regard to this issue