A finest balancing score algorithm to avoid common pitfalls of propensity score matching

03/07/2018

∙

by   Felix Bestehorn, et al.

∙

0

∙

share

Propensity score matching (PSM) is the de-facto standard for estimating causal effects in observational studies. As we show using a case study with 17,427 patients, results obtained using PSM are not reproducible, susceptible to manipulation and partially discard data. We derive four formal properties that an optimal statistical matching algorithm should meet, and propose Finest Balancing Score exact Matching (FBSeM) which meets the aforementioned properties. Furthermore, we prove that PSM will require an exponential bootstrapping effort to obtain a result on par with FBSeM.