BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Stable Weighting and Estimation of conditional treatment effects -
  Ganesh Karapakula (University of Cambridge - Economics)
DTSTART:20230505T143000Z
DTEND:20230505T160000Z
UID:TALK199960@talks.cam.ac.uk
CONTACT:97804
DESCRIPTION:In this paper\, I try to tame "Basu's elephants" (data with ex
 treme selection on observables). I propose new practical large-sample and 
 finite-sample methods for estimating and inferring heterogeneous causal ef
 fects (under unconfoundedness) in the empirically relevant context of limi
 ted overlap. I develop a general principle called "Stable Probability Weig
 hting" (SPW) that can be used as an alternative to the widely used Inverse
  Probability Weighting (IPW) technique\, which relies on strong overlap. I
  show that IPW (or its augmented version)\, when valid\, is a special case
  of the more general SPW (or its doubly robust version)\, which adjusts fo
 r the extremeness of the conditional probabilities of the treatment states
 . The SPW principle can be implemented using several existing large-sample
  parametric\, semiparametric\, and nonparametric procedures for conditiona
 l moment models. In addition\, I provide new finite-sample results that ap
 ply when unconfoundedness is plausible within fine strata. Since IPW estim
 ation relies on the problematic reciprocal of the estimated propensity sco
 re\, I develop a "Finite-Sample Stable Probability Weighting" (FPW) set-es
 timator that is unbiased in a sense. I also propose new finite-sample infe
 rence methods for testing a general class of weak null hypotheses. The ass
 ociated computationally convenient methods\, which can be used to construc
 t valid confidence sets and to bound the finite-sample confidence distribu
 tion\, are of independent interest. My large-sample and finite-sample fram
 eworks extend to the setting of multivalued treatments.
LOCATION:MR12\, Centre for Mathematical Sciences\, Wilberforce Road\, Camb
 ridge
END:VEVENT
END:VCALENDAR