You are not currently logged in.
Access JSTOR through your library or other institution:
A Note on Optimal Matching with Variable Controls Using the Assignment Algorithm
Kewei Ming and Paul R. Rosenbaum
Journal of Computational and Graphical Statistics
Vol. 10, No. 3 (Sep., 2001), pp. 455-463
Published by: Taylor & Francis, Ltd. on behalf of the American Statistical Association, Institute of Mathematical Statistics, and Interface Foundation of America
Stable URL: http://www.jstor.org/stable/1391099
Page Count: 9
Preview not available
The assignment algorithm is an old, well-known, widely implemented, fast, combinatorial algorithm for optimal matching in a bipartite graph. This note proposes a method for using the assignment algorithm to solve the problem of optimal matching with a variable number of controls, in which there is a choice not only of who to select as a control for each treated subject, but also of how many controls to have for each treated subject. The strategy uses multiple copies of treated subjects and sinks with zero cost to absorb extra controls. Also, it is shown that an optimal matching with variable numbers of controls cannot be obtained by starting with an optimal pair matching and adding the closest additional controls. An example involving mortality after surgery in Pennsylvania hospitals is used to illustrate the method.
Journal of Computational and Graphical Statistics © 2001 American Statistical Association