st: nnmatch: Use only some observables for matching, and the others as regression controls?

[Jen Zhen <jenzhen99@gmail.com>]

Dear Statalisters,

I am planning to do some Nearest Neighbor Matching using the -nnmatch-
command, and trying to decide whether to use all of my observable
covariates as matching characteristics, or whether to use only some
and then use the others as regression controls; but I'm not sure what
that choice should depend on?

Now the Stata Journal paper on that command has a section on
Abadie-Imbens' bias-adjustment, using the "biasadj()" option, where it
says that one might want to use for bias adjustment a set of
covariates different from those used as matching criteria; But it does
not hint at when or why one might want to make that choice.

Does anyone have some advice on that?

Many thanks,
J
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/