BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Testing hypotheses via orthogonalization
DTSTART:20260306T151500
DTEND:20260306T161500
DTSTAMP:20260419T233959Z
UID:add42498b8ca7257695ba349c203c39ba8c6082246a6333fd27db3a1
CATEGORIES:Conferences - Seminars
DESCRIPTION:Daniela Witten\, University of Washington\nClassical hypothesi
 s testing frameworks break down in contemporary settings in which null hyp
 otheses are increasingly abstract\, the same data are used to both generat
 e and test hypotheses\, and minimal assumptions about the underlying data 
 are made. In this work\, we propose a new framework for conducting valid h
 ypothesis tests in broad contexts. We propose to add and subtract external
  noise generated from a symmetric shift-family to our data\, X\, to partit
 ion it into two pieces\, X1 and X2. We provide a generic strategy for orth
 ogonalizing X2 against X1 under the null hypothesis H0\, then show that te
 sting whether the orthogonalization was successful provides a valid test o
 f H0 under mild assumptions. Remarkably\, this framework extends naturally
  to the post-selection inference setting with minimal modifications: we si
 mply select a hypothesis on X1\, then perform orthogonalization under the 
 selected null. As our approach neither requires pre-specification of the s
 election mechanism\, nor is it restricted to a small class of data-generat
 ing distributions\, it dramatically expands the settings for which valid p
 ost-selection inference can be conducted. We showcase the flexibility of o
 ur proposal in four case studies: two involving challenging pre-specified 
 null hypotheses and two involving post-selection inference scenarios.\n\nT
 his is joint work with Ameer Dharamshi (University of Washington)\n 
LOCATION:CM 1 517 https://plan.epfl.ch/?room==CM%201%20517
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR