BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Random Graph Asymptotics for Treatment Effect Estimation under Net
 work Interference - Shuangning Li (Stanford University)
DTSTART:20211112T160000Z
DTEND:20211112T170000Z
UID:TALK164227@talks.cam.ac.uk
CONTACT:Qingyuan Zhao
DESCRIPTION:The network interference model for causal inference places all
  experimental units at the vertices of an undirected exposure graph\, such
  that treatment assigned to one unit may affect the outcome of another uni
 t if and only if these two units are connected by an edge. This model has 
 recently gained popularity as means of incorporating interference effects 
 into the Neyman--Rubin potential outcomes framework\; and several authors 
 have considered estimation of various causal targets\, including the direc
 t and indirect effects of treatment. In this paper\, we consider large-sam
 ple asymptotics for treatment effect estimation under network interference
  in a setting where the exposure graph is a random draw from a graphon. Wh
 en targeting the direct effect\, we show that---in our setting---popular e
 stimators are considerably more accurate than existing results suggest\, a
 nd provide a central limit theorem in terms of moments of the graphon. Mea
 nwhile\, when targeting the indirect effect\, we leverage our generative a
 ssumptions to propose a consistent estimator in a setting where no other c
 onsistent estimators are currently available. We also show how our results
  can be used to conduct a practical assessment of the sensitivity of rando
 mized study inference to potential interference effects. Overall\, our res
 ults highlight the promise of random graph asymptotics in understanding th
 e practicality and limits of causal inference under network interference.\
 n\nThis is joint work with Stefan Wager.
LOCATION:https://maths-cam-ac-uk.zoom.us/j/93998865836?pwd=VzVzN1VFQ0xjS3V
 DdlY0enBVckY5dz09
END:VEVENT
END:VCALENDAR