I have a data set where n participants were asked to give a yes or no response to of 12 different (and distinct) stimuli. Each stimulus fits into 1 of 3 categories and an item for each category appears 4 times.

This all fits neatly into a 3x2 contingency table, but each participant is contributing to the table 12 times! I would compare categories with a chi-square test, but the assumption of independence is broken.

Are there alternatives that would account for this? Should I be looking at totaling the responses into 0 to 4 scores?


1 Answer 1


I would try to model this using multivariate logistic regression. This can probably be fit in R with lme4::glmer, some information is at Fitting multivariate linear mixed model in R, follow the links and references therein!

This would need a random intercept for each respondent, shared among the 12 logistic regression, so would really be a multivariate model.


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.