Primary Submission Category: Sensitivity Analysis
Sensitivity analysis for continuous exposures and binary outcomes in matched observational studies
Authors: Jeffrey Zhang, Dylan Small, Siyu Heng,
Presenting Author: Jeffrey Zhang*
Sensitivity analysis frameworks for binary treatments under the matching design have been well-studied, for both continuous and binary responses. Methods for continuous exposures, however, are not well-developed. We discuss how to perform exact randomization inference and sensitivity analysis within the Rosenbaum model with continuous exposures and binary outcomes. We also introduce inference on a generalized notion of the attributable effect. We conduct numerical studies into the design sensitivity and power of various test statistics under several data-generating mechanisms. Finally, we apply our methods to a recent study of the effect of early childhood lead exposure on juvenile delinquency.