BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Harnessing Extra Randomness: Replicability\, Flexibility and Causa
 lity
DTSTART:20231013T151500
DTEND:20231013T170000
DTSTAMP:20260502T031820Z
UID:cb8ca48097009625572bdfe353e27470af5b35d245f53a7bc7a9f68b
CATEGORIES:Conferences - Seminars
DESCRIPTION:Richard Guo\, University of Cambridge\nMany modern statistical
  procedures are randomized in the sense that the output is a random functi
 on of data. For example\, many procedures employ data splitting\, which ra
 ndomly divides the dataset into disjoint parts for separate purposes. Desp
 ite their flexibility and popularity\, data splitting and other constructi
 ons of randomized procedures have obvious drawbacks. First\, two analyses 
 of the same dataset may lead to different results due to the extra randomn
 ess introduced. Second\, randomized procedures typically lose statistical 
 power because the entire sample is not fully utilized.\n\nTo address these
  drawbacks\, in this talk\, I will study how to properly combine the resul
 ts from multiple realizations (such as through multiple data splits) of a 
 randomized procedure. I will introduce rank-transformed subsampling as a g
 eneral method for delivering large sample inference of the combined result
  under minimal assumptions. I will illustrate the method with three applic
 ations: (1) a “hunt-and-test” procedure for detecting cancer subtypes 
 using high-dimensional gene expression data\, (2) testing the hypothesis o
 f no direct effect in a sequentially randomized trial and (3) calibrating 
 cross-fit “double machine learning” confidence intervals. For these pr
 oblems\, our method is able to derandomize and improve power. Moreover\, i
 n contrast to existing approaches for combining p-values\, our method enjo
 ys type-I error control that asymptotically approaches the nominal level. 
 This new development opens up the possibility of designing procedures that
  explicitly randomize and derandomize: extra randomness is introduced to m
 ake the problem easier before being marginalized out.\n\nThis talk is base
 d on joint work with Rajen Shah.\n \nBio: Richard Guo is a research assoc
 iate in the Statistical Laboratory at the University of Cambridge\, mentor
 ed by Prof. Rajen Shah. Previously\, he was the Richard M. Karp Research F
 ellow in the 2022 causality program at the Simons Institute for the Theory
  of Computing. He received his PhD in Statistics from University of Washin
 gton in 2021\, advised by Thomas Richardson. His research interests includ
 e graphical models\, causal inference\, semiparametric methods and replica
 bility of data analysis. Dr. Guo will start as an assistant professor in B
 iostatistics at University of Washington in 2024. 
LOCATION:CM 1 100 https://plan.epfl.ch/?room==CM%201%20100
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR