B1663
Title: Covariate balancing with recurrent marker data
Authors: Kwun Chuen Gary Chan - University of Washington (United States) [presenting]
Raymond Wong - Iowa State University (United States)
Abstract: A nonparametric estimation method is discussed for the counterfactual cumulative rate functions, for a binary exposure. The weights are constructed through balancing covariate distributions between each exposure category and the combined data. Covariate balance is often advocated for objective causal inference since it mimics randomization in observational data. Unlike methods that balance specific moments of covariates, the proposal attains uniform approximate balance for covariate functions in a reproducing-kernel Hilbert space. We will discuss connection between the corresponding infinite-dimensional optimization problem to a finite-dimensional eigenvalue optimization problem. An advantage of such weighting is that it can be applied to multiple functional components in the estimation of cumulative rate functions.