Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 463 - Topics in Biostatistics
Type: Contributed
Date/Time: Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
Sponsor: ENAR
Abstract #313560
Title: A Scanning-Based Inverse Propensity Weighting for Nonrandom Missing Data with Continuous Instrumental Variables
Author(s): David Todem* and Arkaprabha Ganguli
Companies: Michigan State University and Department of Statistics and Probability, MSU
Keywords: Continuous instrument variables; Continuous dichotomizations; Early childhood caries; Empirical process theory ; Exponential tilting, Identifiability; Kernel regression

We consider the problem of estimating unknown population parameters using data with nonrandom missing values, focusing on inverse propensity weighting and instrumental variables. With discrete instruments, Shao and Wang (2016) proposed a semiparametric estimation approach under the framework of exponential tilting propensity. A naive application of this idea to continuous instrumental variables through arbitrary discretizations is apt to be inefficient, and maybe questionable in some settings. In this talk, we propose a novel and flexible approach that does not rely on single arbitrary discretizations but involves continuous dichotomizations across the instrument space. Empirical processes resulting from these dichotomizations are then used to estimate the unknown parameters through weighted integration. We establish the consistency and asymptotically normality of the proposed estimator. Simulation studies and a real data analysis demonstrate the gains of the methodology over procedures that rely on arbitrary discretizations.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program