In a seminar paper I'm discussing all possible matching routines. It turned out that there are quite many of them. Now I am looking for several answers:

First, I discovered that some routines happen to be combined with each other, e. g. Nearest neighbor- and Subclass matching (since there's an extra R-package I do not count e. g. CEM for a 'combination'). Probably there are more combinations applied in practice, or any of my list are also a special case of an other. Could anybody tell me something about that?

Second, are there eventually more routines existing of whose I'm not aware of?

Third, since I'm programming in R I found all the routines in the MatchIt package, except those with the *. Is it possible that these routines are outdated (e. g. there's a package radiusmatching but does not work anymore for R 3.x) or synonyme to an other routine listed?

Here is a list of my findings:

  • Coarsened exact matching (CEM)
  • Full matching
  • Genetic matching
  • Interval matching*
  • Kernel matching*
  • Mahalanobis distance matching
  • Nearest neighbor matching w/ subclasses
  • Nearest neighbor matching w/ replacement
  • Nearest neighbor matching w/ replacement and caliper
  • Optimal matching
  • Propensity score matching*
  • Radius matching*
  • Subclass matching

Note: This post is not intended to discuss dis-/advantages of any routine.

Thanks for any helpful contributions.

  • 2
    $\begingroup$ Since the site has a tag for propensity scores with 199 threads I wonder what steps you have taken to find out about the ones you asterisk? Telling us would help avoid people duplicating your efforts. $\endgroup$
    – mdewey
    Oct 13, 2017 at 13:38
  • $\begingroup$ I want to find out sth. about applied combinations of them, or routines undiscovered from me, or synonymes – just as I stated in my post. PSM is one of 13 routines, for the case I am counting right. Thanks. $\endgroup$
    – jay.sf
    Oct 13, 2017 at 14:07
  • 1
    $\begingroup$ There is also local linear/polynomial matching, and, a bit further afield, inverse probability weighting. You can think of the first as an extension of kernel matching, and of matching as approximation of IPW. $\endgroup$
    – dimitriy
    Oct 13, 2017 at 21:52
  • 1
    $\begingroup$ There's also cran.r-project.org/web/packages/designmatch/index.html $\endgroup$
    – Noah
    Oct 14, 2017 at 23:51
  • 3
    $\begingroup$ There's also Elizabeth Stuart's list. $\endgroup$
    – dimitriy
    Oct 19, 2017 at 5:56


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.