Assistant Professor Jose Zubizarreta, PhD, presented a methodology workshop at the Food and Drug Administration’s 2017 MDEpiNet PPP Annual Meeting.

Zubizarreta’s workshop, titled New Matching Methods for Causal Inference, discussed methodological advancements that overcome the current limitations of standard propensity score matching approaches.  Zubizarreta showed how tools from modern optimization can be used to flexibly balance covariates and handle limited overlap in covariates distributions.  Utilizing the statistical software package “designmatch” for R, Zubizarreta showed how to construct balanced and representative matched samples with binary and multivalued treatment doses.  He also showed how matching methods can be extended for estimation with instrumental variables and in discontinuity designs.  

The workshop discussed the pros and cons of matching methods, and explained how these new methods can improve important aspects of traditional matching approaches.

Enjoy content like this? Sign up for our newsletter and follow HCP on Twitter.