I have collected data from three groups (with different sizes of n). This was done with a test instrument consisting of 9 items, each binary coded (1 = right, 0 = wrong). The results are not equally distributed.

I now want to show that there is a signifikant difference between the three groups for the total score of the test (min = 0, max = 9), but also for the single items.

I did a Kruskal-Wallis for the total score and the results were promising. Now I am unsure if I can use the same test for the individual items because they only have binary values (0 and 1).

So, is it useful to do a Kruskal-Wallis for binary data? Is there any other test I could use?


1 Answer 1


How about something like a chi-squared test for your different groups? You'd have your marginals determined on one axis, but it should be a decent fit for the situation you're describing.


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.