Simplified Population Causal Inference via Balancing

Thursday, June 4, 2015 from 2:00 PM – 3:00 PM

Special Event

Guest Speaker:

Gary Chan, PhD
Associate Professor
Department of Biostatistics and Health Services
University of Washington

 Room W229

The estimation of average treatment effects based on observational data is extremely important in practice and has been studied by generations of statisticians under different frameworks. We will review some common frameworks, such as propensity score, recycled prediction and doubly robust methods, and introduce a model-free method that attains an exact three-way balance of the moments of observed covariates among the treated, the control, and the combined group. We show that efficiency without resorting to direct estimation of propensity score or outcome regression function. Moreover, a non-resampled standard error estimator is proposed and an R package is available. A few practical examples are given to illustrate the practical usefulness of this method.