Hello Jisoo,

You might have a look at GEEs, generalized estimating equations. Here's a link that walks you through the rationale and approach: Article Modeling repeated measures of dichotomous data: Testing whet...

A second approach involves estimating the logistic model via the R library, lmer:

https://stats.stackexchange.com/questions/259055/repeated-measures-with-lmer-and-logistic-regression

A third method follows the SAS proc, catmod:

https://documentation.sas.com/doc/en/pgmsascdc/v_011/statug/statug_catmod_examples09.htm

Finally, for an overview of alternative methods applied to the same RQ/data set, have a look at this link: https://pubmed.ncbi.nlm.nih.gov/8899038/

Good luck with your work.

To my opinion, a Generalized Linear Mixed Model would be the favourite way. It is generalized, because there you can use the binomial distribution for the respone (that's exactly what you need for a dichotomous variable), and the "mixed" part accounts for the correlation of the data within subjects. You can model the condition:time interaction as fixed effects and "subject" as random intercept. It will be possible to extend the model by any number of covariables (e.g. age).

Hello Jisoo Oh . Here is a blog post you may find helpful. I've also given a link to the article the blogger recommends--To GEE or not to GEE! HTH.

• https://www.goldsteinepi.com/blog/multilevelandlongitudinalmodelingapproaches/index.html
• https://journals.lww.com/epidem/Fulltext/2010/07000/To_GEE_or_Not_to_GEE__Comparing_Population_Average.7.aspx