Department of Statistics Seminar
North Carolina State University

presents

Dr. Brett Presnell

University of Florida

"Biased-Bootstrap Recycling"

ABSTRACT

Bootstrap recycling exploits the importance sampling identity to significantly reduce the computational burden in the iterated bootstrap. Unfortunately recycling is directly applicable only in parametric bootstrapping. We show that this limitation can be overcome by embedding the empirical distribution in a nonparametric least favorable family, producing a pseudo-parametric double bootstrap that is amenable to recycling while retaining the desirable properties of the usual nonparametric double bootstrap. As an illustration, we show that Efron's bootstrap percentile interval is second-order correct after calibration by this modified double bootstrap.

Friday, February, 22, 2002

3:35 - 4:35 pm

206 Cox Hall

Refreshments will be served on the second floor of Dabney Hall (left of Room 222) at 3:00 pm.