I ran an experiment where participants were randomly allocated to 2 conditions. They then completed 3 seperate tests with a binary pass/fail outcome for each test.

So I have 3 binary (within-subject) test scores for each participant and 2 (between-subject) groups of participants.

I need to determine:

  1. Were pass rates were equivalent for the 3 tests?
  2. Did the participant's condition influence pass rates overall?
  3. Did condition interact with any particular test outcome?

If the DVs were continous this would be a straightforward repeated measures, mixed model ANOVA.

I have been researching what I should do. I came across Cochran's Q test which seems a perfect way of determining if pass rates varied between the 3 tests but there does not seem to be any way of accounting for the two experimental conditions. Is there a similar test that allows for an additional IV?

I'm using SPSS and have also looked at Generalized Estimating Equations (GEE), which seems like it might be helpful but it appears that there it is only possible to enter a single DV. If I use this procedure how do I enter the 3 repeated test scores?


1 Answer 1


Binomial regression will work for these data. Without going into the technical details (an excellent treatment of these details can be found in Applied Regression Analysis and Generalized Linear Models by John Fox if you wish), this allows you to perform precisely the analysis you alluded to (although it will be dummy-variable regression rather than ANOVA, the results are identical if the model is correctly specified) but using a dichotomous dependent variable. I am aware that this can be performed in SPSS, however, I can offer no specific advice about how, I am most familiar with R.

  • $\begingroup$ Thanks @marcus I think the Generalized Estimating Equations section of SPSS can do this analysis. $\endgroup$
    – mob
    Commented Aug 2, 2012 at 0:41
  • $\begingroup$ Great to hear. You might consider posting a brieft pointer for other SPSS users if you do get it running. Good luck. $\endgroup$ Commented Aug 2, 2012 at 13:09

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.