K-nearest neighbor matching for ATE and missing match.matrix

[lena-222]

Hello guys,

thanks for the great package. I've been using it for matching. However, the K-nearest neighbor matching for ATE would be a good extension for being able to compare different matching algorithms. For the other matching versions the object match.matrix is not available.

Both would be good extensions if it's something you have the capacity for!
Thanks a lot for the good package
Lena

[ngreifer]

Thanks for the suggestion. This is going to be implemented in a major update to MatchIt that hopefully will occur later this year or early next year. Note, though, that standard error estimation for this method is not well understood outside of the context of matching imputation as described by Abadie and Imbens (2006) and implemented in the Matching package. Matching provides an accurate standard error estimator for the ATE when matching with replacement, so for now, I recommend using it instead of MatchIt if you are set on using this method. Otherwise, subclassification, optimal full matching, and generalized full matching can be good alternatives. They don't have a match.matrix because they don't involve create pairs; rather, they form subclasses.

[lena-222]

Hello, thanks a lot for the fast reply! Am Di., 23. Juli 2024 um 21:35 Uhr schrieb Noah Greifer < ***@***.***>:

…

Thanks for the suggestion. This is going to be implemented in a major update to MatchIt that hopefully will occur later this year or early next year. Note, though, that standard error estimation for this method is not well understood outside of the context of matching imputation as described by Abadie and Imbens (2006) and implemented in the Matching package. Matching provides an accurate standard error estimator for the ATE when matching with replacement, so for now, I recommend using it instead of MatchIt if you are set on using this method. Otherwise, subclassification, optimal full matching, and generalized full matching can be good alternatives. They don't have a match.matrix because they don't involve create pairs; rather, they form subclasses. — Reply to this email directly, view it on GitHub <#198 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/APFTKQM6UWHRQ3BPT4SNJ63ZN2V6LAVCNFSM6AAAAABLKIIB2WVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENBWGEZTGNBTGM> . You are receiving this because you authored the thread.Message ID: ***@***.***>

-- Danke und einen wunderschönen Tag! Viele Grüße Lena Schemet