B0971
Title: Debiased nonparametric inference for covariate-adjusted regression functions
Authors: Ted Westling - University of Massachusetts Amherst (United States) [presenting]
Kenta Takatsu - Carnegie Mellon University (United States)
Abstract: The problem of obtaining valid nonparametric inference for a covariated-adjusted (also known as G-computed) regression function with a continuous scalar exposure is discussed. We propose a debiased local linear estimator, and demonstrate that this estimator converges pointwise to a mean-zero normal limit distribution. We use this result to construct asymptotically valid pointwise confidence intervals for function values and differences thereof. Finally, we use recent finite-sample approximation results for the suprema of empirical processes to construct asymptotically valid uniform confidence bands, highlighting, in particular, the technical challenge associated with obtaining faster-than-usual rates of convergence for an empirical process remainder term.
