I want to know how to do a mediation analysis with repeated measures. Here is a fake experiment that has the same form as the one I plan to do (using a fake one lets me simplify some weird complications I already know how to deal with):

For each of 5 targets, 100 participants answer 3 questions:

  1. How warm do you find this person?
  2. How trustworthy do you find this person?
  3. How likely would you be to follow this person's advice.

If there were only 1 target, I would know how to do a mediation analysis to (e.g.,) see whether the relationship between (1) and (3) was mediated by (2). However, I don't know how to do it with 5 targets. Clearly, I should not just average together the scores across all 5 scenarios. And I should not treat each case as different (as though I had 500 participants).

Any suggestions would be hugely appreciated!


1 Answer 1


At least one method for doing this has been published in: Judd et al. Estimating and testing mediation and moderation in within-subject designs. Psychological methods (2001) vol. 6 (2) pp. 115-134 http://www.ncbi.nlm.nih.gov/pubmed/11411437

I haven't yet found a link to code that implements this method though.

  • 4
    $\begingroup$ Could you add a brief summary of this paper? $\endgroup$
    – chl
    May 22, 2014 at 12:43

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.