Multivariate Matching Methods That Are Monotonic Imbalance Bounding

Multivariate Matching Methods That Are Monotonic Imbalance Bounding

March 2011, Vol. 106, No. 493, Theory and Methods | Stefano M. IACUS, Gary KING, and Giuseppe PORRO
The paper introduces a new class of matching methods called Monotonic Imbalance Bounding (MIB), which generalizes and extends the existing Equal Percent Bias Reducing (EPBR) methods. MIB methods aim to improve balance in observational studies by bounding the maximum imbalance between treated and control units, while EPBR methods focus on reducing mean imbalance. The authors provide a detailed review of EPBR methods, define MIB methods, and discuss specific members of the MIB class, including Coarsened Exact Matching (CEM). They demonstrate that MIB methods can improve inferences relative to EPBR methods through analytical results and numerical simulations. The paper also highlights the advantages of MIB methods in terms of model dependence and estimation error, and provides a practical example of how CEM can be used as an MIB method.The paper introduces a new class of matching methods called Monotonic Imbalance Bounding (MIB), which generalizes and extends the existing Equal Percent Bias Reducing (EPBR) methods. MIB methods aim to improve balance in observational studies by bounding the maximum imbalance between treated and control units, while EPBR methods focus on reducing mean imbalance. The authors provide a detailed review of EPBR methods, define MIB methods, and discuss specific members of the MIB class, including Coarsened Exact Matching (CEM). They demonstrate that MIB methods can improve inferences relative to EPBR methods through analytical results and numerical simulations. The paper also highlights the advantages of MIB methods in terms of model dependence and estimation error, and provides a practical example of how CEM can be used as an MIB method.
Reach us at info@study.space
[slides] Multivariate Matching Methods That Are Monotonic Imbalance Bounding | StudySpace