If I have two binary predictors and a binary response how can we calculate the effect size of the interaction term. I use logistic regression from statsmodels package in python.

Will it be ratio of odds ratio (ORR)?

Can I use ORR also for effect size of other interactions like binary-quantitative, binary-ordinal, binary-categorical?

Do we have any guidelines to interpret this term in simple words like small, medium or large (like we can for cohen d)?

  • 1
    $\begingroup$ You should stay away from guidelines, as even Cohen advised against using the ones he designed (see this discussion stats.stackexchange.com/questions/30118/…). The same effect size may be "small" in a given study, but "large" in another. If you don't know if the magnitude of an effect size is important relative to your research question, simply report its value without assigning it a qualifying adjective like "small", "medium" or "large". $\endgroup$
    – J-J-J
    Oct 1, 2023 at 17:48


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.