A New Family of Estimators for Random Walk Problems

In this paper the weighted uniform sampling technique introducedby Powell & Swann (1966) is extended to random walk problemsand generalized. The family so defined includes, as specialcases, not only weighted uniform estimators, but traditionalunbiased importance sampling estimators and an interesting newbiased importance sampling estimator. Model problem analysisshows that biased importance sampling offers the promise ofuniformly low mean square error over a large spectrum of problemtypes.