Importance Sampling for Event Timing Models
by Kay Giesecke of Stanford University, and
October 31, 2011
Abstract: This paper provides an efficient Monte Carlo method for estimating rare-event probabilities in point process models of correlated event timing, which have applications in finance, insurance, engineering, and many other areas. It develops an importance sampling scheme for the tail of the distribution of the total event count at a fixed horizon, and provides conditions guaranteeing the asymptotic optimality of the resulting estimator. The change of measure differs from the widely used exponential twisting. The algorithm applies to point process models with arbitrary stochastic intensity dynamics. Numerical tests illustrate its performance.