I'm using the "Match" package to do propensity score matching.
Here's some
example code that shows the problem that I'm having (much of this code is
taken from the Match package documentation):
*data(lalonde)
glm1 <- glm(treat~age + I(age^2) + educ + I(educ^2) + black +
hisp + married + nodegr + re74 + I(re74^2) + re75 + I(re75^2)
+
u74 + u75, family=binomial, data=lalonde)
X <- glm1$fitted
Y <- lalonde$re78
Tr <- lalonde$treat
# one-to-one matching with replacement (the "M=1" option).
# Estimating the treatment effect on the treated (the "estimand"
option
defaults to ATT).
rr <- Match(Y=Y, Tr=Tr, X=X, M=1);*
And here's where the 'problem' occurs:
*summary(rr) # gives an estimate of 2153.3
mean(rr$mdata$Y[rr$index.treated])-mean(rr$mdata$Y[rr$index.control]) #
gives an estimate of 1083.848
*
Notice that when I simply subtract the means from one another, I get a
different estimate (1083.848) than when the algorithm outputs (2153.3). It
seems that the obvious answer is that I'm not computing the estimate
properly. If so, how is it computed?
One more related question. I'm actually trying to do propensity score
matching to estimate the effect of treatment on a dichotomous variable.
Would the function change at all if the estimated effect is on a
dichotomous scale?
--
Dustin Fife
PhD Student
Quantitative Psychology
University of Oklahoma
[[alternative HTML version deleted]]
