i would use sibling fixed-effects model. it's hard to believe that matching on a few observed characteristics is superior to matching on time-invariant observed and unobserved characteristics. at the very least, you could do both--difference within moms and then match across moms--m 宋时歌 wrote: -- E. Michael Foster getmike.info download vcard: www dot unc dot edu/~emfoster/Foster.vcf Professor of Maternal and Child Health of Health Policy and Administration of Biostatistics (adjunct) Associate Chair for Faculty Advancement (W) 919-966-3773 (F) 919-966-0458 Office: 407C Rosenau Hall << use for FEDEX School of Public Health University of North Carolina, Chapel Hill Rosenau Hall, Campus Box# 7445 Chapel Hill, NC 27599-7445 [the http is throwing off some spam filters) visit the new FAQ section of my blog: fosterfaq dot blogspot dot com 27) [Mollie Ivins] "What do you mean cancer is a gift. You're as a dumb as Shrub. Cancer just killed me. Some gift." "I'm not quite dead yet, sir!" -- Concorde, Monty Python and the Holy Grail 29) Feeling middle-aged? I've got the answer. 8-) - Zelig Mailing List, served by Harvard-MIT Data Center Send messages: zelig(a)lists.gking.harvard.edu [un]subscribe Options: http://lists.gking.harvard.edu/?info=zelig Zelig program information: http://gking.harvard.edu/zelig/

good morning-- this inquiry raised a question i've been meaning to ask involving matching in a multilevel context-- Suppose one has children in a classroom, some of which are exposed to some "treatment" and others are not. i can see two options but would like a third 1) add the classroom identifier as a fixed- or random-effect to the model generating the propensity (if one used the former, there would be some issues with fixed-effect logit, such as having to drop classrooms where all or no kids got the tx) the propensity score would capture the effect of the classroom identifier in the same way it does other covariates 2) forcing the match to exactly match on that covariate (i.e., matching only within classrooms) are there other options? (2) seems too stringent, yet (1) just doesn't seem to capture the qualitative differences among classrooms. For example, we might boost the propensity score for a given case ten percentage points because the treatment is more common in that classroom. the effect would be equivalent to changing the value of some other covariate. i guess that's really what the propensity score really assumes. maybe part of what is bothering me about this scenario is SUTVA and maybe that is contributing to my unease. any one have any insights? regards, michael -- E. Michael Foster getmike.info download vcard: www dot unc dot edu/~emfoster/Foster.vcf Professor of Maternal and Child Health of Health Policy and Administration of Biostatistics (adjunct) Associate Chair for Faculty Advancement (W) 919-966-3773 (F) 919-966-0458 Office: 407C Rosenau Hall << use for FEDEX School of Public Health University of North Carolina, Chapel Hill Rosenau Hall, Campus Box# 7445 Chapel Hill, NC 27599-7445 [the http is throwing off some spam filters) visit the new FAQ section of my blog: fosterfaq dot blogspot dot com 27) [Mollie Ivins] "What do you mean cancer is a gift. You're as a dumb as Shrub. Cancer just killed me. Some gift." "I'm not quite dead yet, sir!" -- Concorde, Monty Python and the Holy Grail 29) Feeling middle-aged? I've got the answer. 8-) - Zelig Mailing List, served by Harvard-MIT Data Center Send messages: zelig(a)lists.gking.harvard.edu [un]subscribe Options: http://lists.gking.harvard.edu/?info=zelig Zelig program information: http://gking.harvard.edu/zelig/