Dear All,
A quick question. I have some data that compares rankings of various
items between different groups. That is, there are ~10 items, each
ranked by 4 different groups.
Can people suggest a suitable test strategy for assessing how similar/
dissimilar the rankings are. I think a sensible null hypothesis would be
to use the lumped ranking as the null hypothesis, and then test for
difference from that. Using a null hypothesis that there is no ranking
seems not as useful (I'm interested in differences in ranking, not
whether there /is/ a ranking).
I'm also not sure which test to use - Can one generalise (e.g.)
Spearman's/ Kendall's to multiple rankings?
Thanks,
Matt
Dr. M. Williams
Clin. Onc SpR.
Clinical Research Fellow, CRUK
|