Repeated measures analysis with ordinal data

#1
Hello,

I have data on subjects' ratings of 5 potential side effects (rated as: none, mild, moderate, considerable, strong) following three different interventions (one placebo, and two active) --which were administered in a randomized manner, each 3 weeks apart (within-subjects design).

I would like to determine whether the side effects reported for either of the two active interventions differed from those reported for the placebo intervention.

I considered a chi-square analysis whereby the reports for the placebo condition would be the 'expected' and the reports for an active intervention would be the 'observed.' But the within-subjects/repeated measures design violates the Chi-square assumption of independence.

Should I be using some form of non-parametric repeated measures ANOVA?
(I'm carrying out the analyses in SPSS, btw)

Any help/advice would be much appreciated.

Best,
Siobhan
 
#2
Hello Siobh,

I'm currently dealing with a similar problem. It seems that you would need to perform the Scheirer Ray Hare Test, which is the two-factor equivalent of the Kruskal-Wallis test.

To the best of my knowledge, this is not possible in SPSS. However, it seems that running a two-factor ANOVA with the rank scores of the ordinal data would yield the same results. I can't confirm that, though. Maybe someone with more experience can tell us =).

Good luck!