Directional Penalties for Optimal Matching in Observational
Studies
Description
Improves the balance of optimal matching with near-fine balance by giving penalties on the unbalanced covariates with the unbalanced directions. Many directional penalties can also be viewed as Lagrange multipliers, pushing a matched sample in the direction of satisfying a linear constraint that would not be satisfied without penalization.
Yu and Rosenbaum (2019) .