Dear MatchIt Mailing List,
when I calculated some summary statistics to compare means before and after
matching, I used the standardized = TRUE option to obtain standardized mean
differences. However, to my surprise, the output shows that the standard
deviation in the control group is used for standardization (where I felt
the SD of the treatment group was more appropriate but I can live with
that). You can produce the output I am referring to via:
summary(matchit(formula = treat~re74+re75+ black+hispan+educ+age, data =
lalonde), standardized = TRUE)
As I was unsure which SD was used to standardize, I dug in the code and
noticed, that in the MatchIt:::qoi function the SD is actually calculated
from the treatment group, i.e. I am almost sure that in the code of qoi,
the SD of the treated is calculated.
I am using a very recently built version of Matchit from CRAN but the
underlying code seems not to have been updated since 2013. Has this been
addressed/fixed in the meantime?
With kind wishes,
Philipp Doebler
Version information:
Package: MatchIt
Version: 2.4-21
Date: 2013-06-26
Date/Publication: 2013-06-28 00:55:57
Built: R 3.2.0; ; 2015-05-18 19:26:11 UTC; unix
Dear List good morning
I have a question
How is the standardize mean difference in full matching method for both continuous and categorical
variables? In unmatched sample if I use like standardized mean difference for categorical variables this formulae:
m <- pcontrol-ptrt # where p stay for percentual
a<- ptrt*(1-ptrt) + pcontrol*(1-pcontrol)
z <- m/sqrt(a/2)
the results match with matchit...
but I' m not able to caluculate standardized mean difference into full matched sample where there are the weights
Can You so gentle to help me
Thank You in advance
Bonitta Gianluca
Dear list
Is there a way to estimate the ATT and ATE effect sizes after calling MatchIt? I am using MatchIt to perform optimal matching for causal inference.
Thanks in advance
Barth Riley