Resampling in the Undergraduate Curriculum

• Activity Number: 276
• Type: Roundtables
• Date/Time: Tuesday, August 11, 2015 : 7:00 AM to 8:15 AM
• Sponsor: Section on Statistical Education
• Author(s): Tim Hesterberg*
• Companies: Google
• Keywords: education ; bootstrap ; permutation test ; simulation

Abstract

Let's talk about resampling! Bootstrapping and permutation tests have great potential benefit for helping students understand hard concepts---sampling distributions, p-values, and standard errors---by giving them concrete pictures using tools they're familiar with, like histograms. And in many cases, these methods are easier and more accurate than classical approximations. We use them all the time at Google :-) But there are some subtleties, things at least the instructor should know. For example, the usual bootstrap standard errors are too small, and the simplest bootstrap confidence intervals under-cover in small samples and do exactly the wrong thing for biased statistics. I'll share thoughts on how to approach resampling for the introductory and junior-level "mathematical statistics" courses. I'll also discuss how resampling fits into the revised ASA Guidelines for the Undergraduate Curriculum.