Hi Shane,
To estimate the ATE following full matching you would need to create new weights yourself,
that would weight the subjects in each subclass up to the combined treatment/control
sample (right now the matchit weights weight them to look like the treatment group, thus
estimating the ATT). The way the matchit weights are created is detailed here, so you
could potentially modify that to create your own weights:
http://gking.harvard.edu/node/4355/rbuild_documentation/How_Exactly_are.html
Depending on exactly how it is done, optimal matching is generally just a 1:1 matching
algorithm that matches control subjects to treated subjects (discarding unmatched
controls) and so you generally can't estimate the ATE using that approach.
Liz
On 9/19/11 8:10 PM, "Shane Phillips" <phillips.shane(a)gmail.com> wrote:
Hi!
How would I go about estimating ATE in R following full matching and optimal matching in
MatchIt? Would it be any different than the method shown in the documentation for nearest
neighbor matching?
Thanks!
Shane