Post date: Dec 5, 2017 5:29:18 PM
Causal & SLAM Working Groups Seminar and Short Course
TIME: 11/30/2017
Thursday 12:00-1:30pm (lunch provided)
Place: Room W3011
Title: Title: Complier classification with sharp instrumental variables
Speaker: Edward Kennedy, Department of Statistics, Carnegie Mellon University
Abstract: It is well-known that, without restricting treatment effect heterogeneity, instrumental variable (IV) methods only identify ``local'' effects among compliers, i.e., those subjects who take treatment only when encouraged by the IV. Local effects are controversial since they seem to only apply to an unidentified subgroup; this has led many to denounce these effects as having little policy relevance. However, we show that such pessimism is not always warranted: it can be possible to accurately predict who compliers are, and then nearly identify effects in observable subgroups. We first discuss the problem of guessing compliance status using covariates, and show that the error rate in doing so can be arbitrarily small, even for very weak IVs. We go on to propose a new measure of IV quality called ``sharpness'', which is completely separate from strength, and reflects the variation in compliance explained by covariates. We show that sharper IVs allow better complier identification, and yield tighter bounds on effects in observable subgroups. We also propose nonparametric efficient estimators of sharpness and corresponding effect bounds, which can incorporate modern machine learning tools. Finally we apply the methods in two studies, one where the IV is strong but blunt, and another where the IV is weak but sharp.
TIME: 12/1/2017
Place: Genome Café (Room E3609)
On Friday (Dec. 1) Edward will give an all-day tutorial about empirical process theory and semiparametric inference (based on these papers). The tutorial, which will also take place in the Genome Cafe, is scheduled as follows:
10:30-12 Morning Session
12-1:30 Lunch (provided)
1:30-3 Afternoon Session