Efficient simulation for dependent rare events with applications to extremes (Q1703036)

Let \((X_1,\dots,X_n)\) be a random vector and \(M=\max_i\,X_i\). The authors are interested in the estimation of \(\alpha(\gamma)=P(M>\gamma)\) using estimators based on \(E(\gamma)=\sum_{i=1}^{n}I_{[X_i>\gamma]}\). Two estimates are proposed and their properties studied. Aside from that, the authors are also interested in the estimation of \(\beta_n(\gamma)=E\,\big[Y\,I_{[E(\gamma\geq n)]}\big]\), \(n=1,2,\dots,d,\) where~\(Y\) is an arbitrary random variable. No assumptions about the dependence of the events \([X_i>\gamma]\) or dependence of these events and random variable~\(Y\) are done. Properties of the suggested estimators are studies together with their efficiency. The numerical performance is illustrated on several nontrivial examples.