Biostatistics and Medical Informatics Seminar with Fei Zou (UNC)

Dr. Zou will introduce deepHTL (deep Heterogeneous Treatment Learning), a semiparametric framework for testing and estimating heterogeneous treatment effects (HTE) using observational data where complex confounding exists. deepHTL incorporates three key components to deliver robust causal inference: bagged deep neural networks; bias-reduces semiparametric regression; and global kernal score testing based on the Davis method and a cross-fitted permutation test.