San Francisco Bay Area Chapter of American Statistical
Association (SFASA) Virtual Seminar
February 19th, 2021
Topic
Speaker
Time/Date
Registration Access
2:00 – 3:00pm PST,
Friday, February 19th,
2021
Link to the free registration site
Access info
will be sent to registered attendees via e-mail
Abstract
Robust
Estimation of Heterogeneous Treatment Effects using Electronic Health Record
Data
Dr.
Honglang Wang - Indiana University–Purdue University Indianapolis
Causal
estimation of the heterogeneous treatment effects is an essential component of
precision medicine. Different estimators have been developed under various
situations to accommodate specific studies' practical needs to achieve desired
analytical properties. In this presentation, I describe a general framework
that unifies many existing estimation methods through a common formulation of
score equations. This formulation covers many of the frequently used causal
estimators, including the inverse propensity weighting, augmented inverse
propensity weighting, R-learning or efficient A-learning, and modified
covariate methods with and without efficiency augmentation. The proposed
framework also allows for more flexible choices of loss functions and thus
enhancing the estimator's robustness against data irregularities in
observational studies. Asymptotic properties of the general estimator have been
developed so that valid inferences can be made. As an application, data from a
local electronic health record system are analyzed to determine the causal
effects of two antihypertensive therapies while accounting for patient
characteristics.
About
the Speaker:
Dr.
Honglang Wang is an Assistant Professor of Statistics in the Department of
Mathematical Sciences at IUPUI. He received his PhD in Statistics and dual PhD
in Quantitative Biology from Michigan State University in 2015. Dr. Wang’s
research concerns statistical methods for longitudinal/functional data, high
dimensional data, causal inference, empirical likelihood, gene environment
interaction, and machine learning.