I'm trying to examine the difference in proportions of outcomes from three different tests (positive, negative, unsure). I want to compare what proportions of outcomes is positive for each test. All three tests are performed on the same group of 100 subjects (so each subject underwent three different tests). I figured a McNemar test is most suitable as the results are correlated (because the same subjects were tested), cathegorical and I want to examine the proportions. Each test yields a completely different range of outcomes so the categories 'positive', 'negative' and 'unsure'are assigned differently for each test. This makes me think i can't use a Cochran Q-test, but I correct me if i'm wrong.

My question is how to do a McNemar test for three treatments in R. I've been looking far and wide on the internet but can't seem to find an answer. This article suggests bootstrapping to compensate for the correlation but i thought that was already accounted for in a McNemar test.


  • $\begingroup$ I don't understand your main goal: "I want to compare what proportions of outcomes is positive for each test". Could you try to specify the null and alternative hypothesis behind your question? $\endgroup$ – Michael M Apr 25 '17 at 11:40
  • $\begingroup$ My null hypothesis is that all the tests yield the same proportion of positive subjects, the alternative is that they don't. $\endgroup$ – Zip Apr 25 '17 at 17:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.