Dear all,
I am using -nnmatch- with the bias adjustment option and have difficulties to figure out how to calculate the potential outcomes using the matching dataset that -nnmatch- produces.
In the example below, the matching dataset "d1" contains further variables labeled "bias adjustment variables":
-age_0b educ_0b black_0b hisp_0b married_0b re74_0b re75_0b u74_0b u75_0b age_1b educ_1b black_1b hisp_1b married_1b re74_1b re75_1b u74_1b u75_1b-
How do I involve these "bias adjustment variables" to get the correct predicted potential outcome for the control group "re78_0a" and thus the same estimated ATT as in the output of -nnmatch-?
Thanks, David.
I am using -nnmatch- with the bias adjustment option and have difficulties to figure out how to calculate the potential outcomes using the matching dataset that -nnmatch- produces.
In the example below, the matching dataset "d1" contains further variables labeled "bias adjustment variables":
-age_0b educ_0b black_0b hisp_0b married_0b re74_0b re75_0b u74_0b u75_0b age_1b educ_1b black_1b hisp_1b married_1b re74_1b re75_1b u74_1b u75_1b-
How do I involve these "bias adjustment variables" to get the correct predicted potential outcome for the control group "re78_0a" and thus the same estimated ATT as in the output of -nnmatch-?
Code:
* Dataset from:
* Abadie, Alberto, David Drukker, Jane Leber Herr, and Guido W. Imbens, 2004,
* Implementing matching estimators for average treatment effects in Stata,
* The Stata Journal 4, 290–311.
use "https://github.com/gvegayon/nnmatch2/blob/master/ldw_exper.dta?raw=true", clear
* The bias-corrected estimator
nnmatch re78 t age educ black hisp married re74 re75 u74 u75, ///
tc(att) m(1) metric(maha) bias(age educ black hisp married re74 re75 u74 ///
u75) keep(d1) replace
* Estimated ATT = 1.533125
use d1
bysort id: egen re78_0a = mean(re78_0)
gen diffOutcomes = re78_1 - re78_0a
egen tag_id = tag(id)
sum diffOutcomes if tag_id
* (My) estmiated ATT = 1.554697 (which is the value for "no bias adjustment")
