Matchit October 2013

matchit@lists.gking.harvard.edu

8 participants
4 discussions

by Sharon Simonton

Hello, I've been trying to print out (or export) a data set containing the IDs for my matched pairs to read them into SAS/Mplus. I just realized that what I've been doing is not working correctly. I've been using the syntax example in the manual given for seeing the outcomes for matched pairs to try to print out the IDs for my matched pairs. cbind (dataset [row.names (m.matchset$match.matrix), "I_ID"], dataset [m.matchset$match.matrix, "I_ID]) This doesn't seem to be giving me what I want. How can I get IDs and a unique pair number for my matched pairs so that I can do conditional logistic regression? Thank you, Sharon Simonton

10 years, 6 months

Two small points/issues: handling NAs and non-forula formulas

by Matthieu Stigler

Hi Thanks a lot for the MatchIt package, very useful! I encountered two small issues while using it (code to reproduce at the end of e-mail), maybe these could be made more clear in the code? 1 handling NAs As already reported (long time ago, see: https://lists.gking.harvard.edu/pipermail/matchit/2006-April/000364.html ) matchit requires all the columns in the data to have no NAs, even if these columns are not used in the match/estimation. Is this really necessary? Maybe could be made more tolerant by checking for NAs only in the variable used? Maybe the line: if (sum(is.na(data)) > 0) could be changed into something like: if(any(is.na(model.matrix(formula, data=data)))) 2 Formula General functions in R accept "character formula", such as: lm("y~price.index", data=freeny) However, matchit doesn't: matchit("y~price.index", data=freeny) maybe this could be checked, and either corrected, or warned (the error message "$ operator is invalid for atomic vectors" is not very helpful to find out the problem..)? Like: if(!is.formula(formula)) formula <- as.formula(formula) or: if(!is.formula(formula)) stop("Argument formula should be of class 'formula' ") Illustrative code: library(MatchIt) data(lalonde) m.out_exact <- matchit(treat ~ educ + black + hispan, data = lalonde,method = "exact") lalonde2 <- lalonde lalonde2[1,"married"] <- NA m.out_exact <- matchit(formula=treat ~ educ + black + hispan, data = lalonde2, method = "exact") ## Formula issue? form <- "treat ~ educ + black + hispan" glm(form, data=lalonde) m.out_exact <- matchit(form, data = lalonde,method = "exact")

10 years, 6 months

Concatenating matched data sets to look at balance diagnostics for pooled data set

by Sharon Simonton

Hi All, I've matched using a pattern mixture model approach to address missing data for several of the covariates used in my propensity score model. I now have four separate MatchIt output files of matched pairs for separate patterns for missing data. I'd like to concatenate them so that I can look at the balance statistics/diagnostics for the full pooled set of matched pairs that I've been using for analyses (have balance statistics for each individual data set). Is there a straightforward way to do this? Alternatively, I actually already have a pooled data set with all of the pairs and associated measures. Is there a way to import this into MatchIt to get the balance statistics? Thank you for your help. MatchIt is terrific! Best wishes, Sharon Simonton University of Michigan

10 years, 6 months

External propensity scores

by Andrea Lamont

Hello: I am running a model with propensity score matching. My data had a great deal of missingness - including missingness on the matching variable - so I used multiple imputation prior to matching. To avoid having duplicate cases matched (as result of having 10 imputations per observation), I ran a logistic regression (external to matchIt) predicting membership, and averaged the resultant fitted values for a final propensity score. I would now like to use this score as the propensity score in MatchIt. The problem is that I don't really have a formula. Is there a way to read this external propensity into MatchIt and match on that variable? Thanks, Andrea -- Andrea Lamont, MA Clinical-Community Psychology University of South Carolina Barnwell College Columbia, SC 29208 Please consider the environment before printing this email. CONFIDENTIAL: This transmission is intended for the use of the individual(s) or entity to which it is addressed, and may contain information that is privileged, confidential, and exempt from disclosure under applicable law. Should the reader of this message not be the intended recipient(s), you are hereby notified that any dissemination, distribution, or copying of this communication is strictly prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy/delete all copies of the original message.

10 years, 7 months

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Matchit October 2013