, because its sample variance estimator is an inefficient estimator of the true variance. We propose a simple remedy: to still … completely new sequential importance sampling estimator of the desired tail probability. Numerical experiments suggest that the … sequential importance sampling estimator can be significantly more efficient than its competitor. …

from the original populationas replacements for units who have dropped out of the panel, aso--called refreshment sample …