Using Sectioning to Construct Confidence Intervals for Quantiles When Applying Antithetic Variates

摘要

We use sectioning to construct a confidence interval (CI) for a quantile when applying antithetic variates, which is a well-known variance-reduction technique. Closely related to batching, sectioning starts with the batching CI and replaces the batching quantile estimator with the overall quantile estimator throughout. Since the overall quantile estimator often has smaller bias than the batching quantile estimator, the sectioning CI is centered at a better point estimator, and this should lead to sectioning CIs having better coverage than batching CIs.