How do I do McNemar test for propensity matched data in R?

Question

I did propensity score matching for retrospective data, and now I want to run outcome analysis. For dichotomous variables, I heard that I need to do McNemar test since I should treat propensity score matched data as paired ones. Typical t-test table for dichotomous variables would be like:

	Treatment group	Control group
Yes	number	number
No	number	number

However, for McNemar test, I figured out that I need another vesion of table like:

	Treatment group Yes	No
Control group Yes	number	number
No	number	number

But I can't figure out how to find those numbers in table above. Because there are a lot of outcome variables, and number of participants in each group, I cannot count each pair as well. Please tell me how I can make a table needed and run McNemar's test in R.

Answer 1

McNemar's test isn't the only way to test the association between treatment and a binary outcome in matched data. You can simply run a logistic regression of the outcome on treatment and use a cluster-robust standard error that accounts for the paired nature of the data. This procedure is recommended by Abadie and Spiess (2022) who prove that it is valid. This is explained in MatchIt's estimating effects vignette. There are a number of benefits to using a regression framework: you can include covariates in the model, you can compute a specific quantity rather than just a p-value, and you can compute a confidence interval around that quantity.