I am looking for a non-parametric version of the multivariate repeated measures GLM procedure.

I have ordered symptoms of a disease at 2 time points (0-3), and I want to show that they change differentially (in response to something that happens to the whole sample during the 2 time points). If my symptoms were metric, I could run a multivariate repeated measures GLM, and if the interaction term symptom*time is significant, I have differential change of symptoms over time.

The Friedman test - a non-parametric version of the repeated measure GLM - does not suffice, because it can either compare medians of multiple DVs at the same time, or medians of the same DV over time, but it cannot handle multivariate + repeated measure (at least I haven't found a way to do that).

Software I have installed and can work with is R, SPSS, and MPLUS available. Thank you.

  • $\begingroup$ Wouldn't ordered logit be what you are looking for? $\endgroup$
    – January
    Jul 5 '13 at 11:25
  • $\begingroup$ Ordered logit might be the link function, but in what software can I perform an ordered logit multivariate repeated measure analysis? $\endgroup$
    – Torvon
    Jul 5 '13 at 12:17
  • 1
    $\begingroup$ Would this: cran.r-project.org/web/packages/repolr/repolr.pdf be your answer? I have never worked with it, but it looks as it might, pardon the pun, fit to your requirements. $\endgroup$
    – January
    Jul 5 '13 at 12:31
  • $\begingroup$ It allows modeling one ordered DV over 2 time points (which you can do with 3 clicks in SPSS via Friedman test), but not of several DVs over 2 time points :( $\endgroup$
    – Torvon
    Jul 5 '13 at 12:57
  • $\begingroup$ Aren't you looking for a generalized linear mixed effects model (i.e., glmer in the lme4 package)? cran.r-project.org/web/packages/lme4/lme4.pdf $\endgroup$ Jul 5 '13 at 14:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.