A0735
Title: Robust and efficient nonparametric estimation of transported total causal effects and mediation causal effects
Authors: Kara Rudolph - Columbia University (United States) [presenting]
Ivan Diaz - NYU Langone Health (United States)
Abstract: The focus is on identifying and estimating transported causal effects in the context of encouragement-design interventions---those that encourage uptake of exposure of interest. We consider types of transported total effects, like the intent-to-treat average treatment effect, the average treatment effect of the exposure, and the complier average treatment effect. We also consider transported interventional indirect and direct effects of the encouragement on the outcome that operate through mediators or not, respectively. We describe robust and efficient estimators for each type of transported effect and apply them to motivating research questions from the Moving to Opportunity Study (MTO). MTO is a large-scale encouragement-design intervention in which Section 8 housing vouchers, which encourage families in public housing to move by subsidizing rents on the private market, were randomly assigned. In the context of MTO, we use the transport estimators in service of understanding the reasons for differences in site-specific effect estimates. These transport estimators may also be useful in other scenarios-including, to generate place-specific intervention effect estimates, in problems related to surrogacy, or in other data-fusion-related problems. We end with current work on extending these identification results and estimators to accommodate more general data structures.
