Does importance sampling overestimate? I mean, because we are putting more weight on a particular area of the distribution (say the tail), would such techniques not overestimate tail probabilities?
