Title: New matching methods for causal inference using integer programming
Authors: Jose Zubizarreta - Harvard University (United States) [presenting]
Magdalena Bennett - Columbia University (United States)
David Hirshberg - Columbia University (United States)
Juan Pablo Vielma - MIT (United States)
Abstract: In observational studies of causal effects, matching methods are often used to approximate the ideal study that would be conducted if it were possible to do it by controlled experimentation. We will discuss new matching methods based on integer programming that allow the investigator to overcome three limitations of standard matching approaches by: (i) directly obtaining flexible forms of covariate balance; (ii) producing self-weighting matched samples that are representative by design; and (iii) handling multiple treatment doses without resorting to a generalization of the propensity score. (iv) Unlike standard matching approaches, with these new matching methods typical estimators are root-n consistent under the usual conditions. We will illustrate the performance of these methods in real and simulated data sets.