Hi Sharon,
Yes, I agree with what you write below and I think you have the right idea. You can use
code like what I have below to send MatchIt your propensity scores, and those will be what
MatchIt will use to implement the matching/subclassification/whatever.
Good luck with it!
Liz
On Jul 23, 2013, at 4:51 PM, Sharon Simonton
<simonton@umich.edu<mailto:simonton@umich.edu>> wrote:
Hi Liz,
Thank you so much for your response! I've been thinking this through a bit more. The
pattern mixture models will allow me to estimate a propensity score for respondents having
missing data on the covariates outside of MatchIt (using Mplus or SAS). I'll then
read this into MatchIt and will put a missing code into my covariates having missing data
so that the matching methods in MatchIt will run.
I think that MatchIt will then allow me to use all of the matching methods based on the
propensity score (basically all except Mahalanobis matching and CEM that actually use the
raw data for matching). Missing codes of any kind shouldn't interfere with or affect
this since I won't be using MatchIt to estimate the propensity scores. Am I correct
in assuming that MatchIt would only use the covariate values for computing the balance
statistics in such a situation? I think that I should still be able to use a stratifying
variable with complete data for matching if need be (e.g. a categorical variable
indicating groups of subjects having complete data and groups having different patterns of
missingness)
My balance statistics won't be meaningful --unless I can figure out some way to
recode my data to make this work. But I will have matched pairs/sets via several different
methods ready for analyses!
I'm going to try to figure out if I can find a way to recode the covariates to use
MatchIt to assess balance. But if this doesn't work, I think I can find a way to do
this with another software package. Sound reasonable?
Thanks again for your help!
Sharon
On Mon, Jul 22, 2013 at 9:06 PM, Stuart, Elizabeth A.
<estuart@jhsph.edu<mailto:estuart@jhsph.edu>> wrote:
Hi Sharon,
On your first question I don't know of an easy way to integrate a pattern mixture
model with MatchIt, but yes, you could estimate the propensity score outside of MatchIt
and then send that to MatchIt for the matching. You could also check balance on missing
data indicators by using the "addlvariables" option in the summary() command.
http://r.iq.harvard.edu/docs/matchit/2.4-20/Arguments2.html
And I don't follow your 2nd question fully because the link didn't work for me,
but are you trying to ask about this FAQ on how to send matchit your own propensity
score?
http://r.iq.harvard.edu/docs/matchit/2.4-20/How_Do_I3.html
You can send MatchIt your own propensity score, you just use code like:
m.out <- matchit(treat ~ x1 + x2 + x3, data=dta, distance=dta$pscore)
It uses the formula treat ~ x1 + x2 + x3 only to figure out which variables to do the
balance diagnostics on; the matching is done using the variable dta$pscore.
I hope this helps!
Liz
On Jul 22, 2013, at 11:51 AM, Sharon Simonton
<simonton@umich.edu<mailto:simonton@umich.edu>> wrote:
Dear All,
Thank you for a great software package! I'm currently working on a project where MI
is not a viable option. I'm hoping to use pattern mixture models to estimate
propensity scores for cases and controls having missing data for covariates. Is there then
any way to then use MatchIt for matching and computing balance statistics? Hoping that
there might be a way to use stratification or missing data indicators to do this (also
considering general location models-- just in case these would work better here).
I'll be tremendously appreciative of any thoughts or advice you may have on this.
Also want to let you know that I can't open the documents that you show links to in
your responses (e.g
http://gking.harvard.edu/node/4355/rbuild_documentation/How_Do_I3.html.<…
Figured out the syntax here but would be helpful to be able to see the docs you reference.
Is there anything I can do to access them?
Thank you!
Best wishes,
Sharon Simonton
University of Michigan
Ann Arbor, MI
-
---
MatchIt mailing list served by HUIT
List Address:
matchit@lists.gking.harvard.edu<mailto:matchit@lists.gking.harvard.edu>
Subscribe/Unsubscribe:
http://lists.gking.harvard.edu/mailman/listinfo/ei
MatchIt Software and Documentation:
http://gking.harvard.edu/matchit/
Browse/Search List Archive:
http://lists.gking.harvard.edu/mailman/private/matchit/
Matchit mailing list
Matchit@lists.gking.harvard.edu<mailto:Matchit@lists.gking.harvard.edu>
To unsubscribe from this list or get other information:
https://lists.gking.harvard.edu/mailman/listinfo/matchit